Journal of Computing & Biomedical Informatics

Academic Advancement: Libraries' Integral Support for Research in Public Sector Universities of Khyber Pakhtunkhwa, Pakistan

Tahira Bibi — 2024-06-01

This research article explores into the pivotal role of libraries in enhancing research activities within the public sector universities of Khyber Pakhtunkhwa, Pakistan. The study's objectives encompassed assessing current research services, evaluating technology integration, and examining collaborative spaces within these libraries. A quantitative research approach was employed, surveying eighty library professionals across thirty-four universities. Findings revealed a diversified demographic profile among respondents, emphasizing the need for tailored research support. Analysis of digital resources highlighted moderate satisfaction with research databases but a positive perception of data management services. Technology integration, including interlibrary loan systems and institutional research repositories, showcased promising levels of effectiveness. Collaborative spaces within libraries, along with IT support and research advisory services, were deemed valuable for research endeavors.Recommendations emerged to enhance digital resources' effectiveness, optimize technology integration, and expand collaborative spaces. Improving IT support, empowering research advisory services, and promoting research outreach initiatives were also suggested. Continuous evaluation and adaptation mechanisms were emphasized to create a dynamic and effective research support ecosystem within university libraries. This study contributes essential insights into enhancing research support services, fostering collaboration, and promoting scholarly excellence within university libraries in Khyber Pakhtunkhwa, Pakistan.

The Impact of Virtual Reality (VR) Applications in Ming-style Furniture Designs

Gan Saixiong — 2024-06-01

The research investigates how VR applications can radically change Ming-style furniture design. It explores the complex factors that control how satisfied a person is with furniture designed through VR. The purpose of this study is to understand the complex effects of VR technologies on the creative process and consumer instincts through the DoE research methodology, as well as the experience level. Utilizing VR applications in the design of Ming-style furniture sparks important inquiries concerning the effectiveness, practicality, and attraction of virtual design surroundings. To enhancing the processes and effects of designing, it is essential to appreciate how design method, acquaintance level, furniture captiousness, and the mechanism of additional viewpoint that influences the satisfaction of designers, so as to fully utilize the VR technology. 385 individuals who used Ming-style furniture in Suzhou completed a survey over the Internet, which contributed significant insights into VR-mediated design experiences. By determining satisfaction perception, collecting data via Likert scales, and then using SPSS, core-region data is processed, which resulted in examining the correlation of multi-regression among the variables. The results show that there is a strong connection between Virtual Reality Implementation and User satisfaction of Ming style. This investigation emphasizes the great importance of high-quality ways to receive feedback and users who are experienced in how they make users happy while using virtual and real-world ways to communicate means of doing or reaching something and how skillful they are at creating visual images in the mind of Ming-style furniture design. By clarifying the numerous forces that underline what users see, feel, and learn from a product line, this work makes more likely the discovery and spread of counter-intuitive products for everyone.

Methodology for Ensuring Secure Disease Prediction using Machine Learning Techniques

Aleena Imran — 2024-06-01

In today's digital world, the e-healthcare system has increased the patient data in vast amount. Protecting this confidential patient data from unauthorized access and tampering is crucial as the data contains sensitive details regarding patient’s health and any tampering on such details would result in manipulation of patient data which could lead to misdiagnosis and incorrect treatment plans. Conventional healthcare systems lack the ability to secure patient data from unauthorized access which eventually leads to data tampering and data loss. Data security and data privacy are crucial components within the healthcare sector and can be enhanced by the utilization of blockchain framework. Within the healthcare domain, disease identification and prediction is also a critical challenge. This study focuses on disease detection and prediction such as diabetes mellitus and blood pressure by implementing ML models such as Decision Tree, SVM, KNN, Naive Bayes, Random Forest and ensemble learning while maintaining the integrity of patient sensitive health data. The diagnostic results predicted by classifier and new patient data have been stored on smart contracts. Only authorized persons such as healthcare professionals can have access to patient sensitive health related data and diagnostic results predicted by the classifier. The research aims to enhance the efficiency of machine learning classifiers along with data integrity.

A Genetic Algorithm for Targeted Regression Testing: Balancing Coverage and Change Focus

Fahad Nazir — 2024-06-01

The selection of regression test cases is used to choose a subset of test suits that are used to exercise the altered program to ensure that the modified part has no unintended consequences on the unmodified part of the program. In previous works, the single objective is used for the selection of test cases. In this thesis, we preserved test case selection as a multi-objective optimization problem. We select code coverage and code change information for test case selection. There are many different methods for test case assortment (i.e., firewall, textual differing g, test-tube, etc.). We used a genetic procedure for multi-objective test case assortment. Our proposed technique first collects the related information such as the size of the system under test, the size of the test suit, and modification between the original and modified versions of the system. Then collect and analyze the code coverage and code change information, also collect user requirements for test case selection. Finally, select a subset of test cases based on cost, fault uncovering, code coverage, and code alteration information for test case selection. The proposed system is for a moderate-level desktop application. Three datasets (Triangle, Tree data structure, and Jodatime) are used for the experiment of the proposed system. For the evaluation of the proposed test selection technique, we used precision and recall evaluation matrices. Our experiment study demonstrates that our proposed technique selects almost 75% of related test cases.

Enhancing the Inverted Topp-Leone Geometric Distribution: Properties and Applications in Computing

Saadia Tariq — 2024-06-01

Statistics has developed a substantial interest in lifetime models, specifically within the domain of statistical inference. Practical domains such as computer science, medicine, engineering, biology science, management, and public health make extensive use of these models. Probability models find application in various domains, including game-winner prediction, team classification, winning margin evaluation, and likelihood of team victory. Recently, the model of mixed distribution has gained widespread recognition in the field of statistical data modelling. This paper aims a new two-parameter generalization of inverted Topp-Leone distribution. The new model, known as the Inverted Topp-Leone Geometric distribution, is created by mixing inverted Topp-Leone and geometric distributions. The quantile function, incomplete moments, ordinary moments, median, mode, mean residual life function, entropy, Shannon entropy, and mean deviation are some of the mathematical features of the new distribution that are obtained. Other characteristics include the mean deviation. The maximum likelihood approach is used to arrive at an estimate of the parameters of the model. Inverted (or inverse) distributions are advantageous for investigating further characteristics of the phenomenon. The behaviour of the parameter estimations is investigated by a Monte Carlo simulation. A practical computing application is provided to illustrate the new model's usefulness.

Malaria Cell Classification through Exercising Deep Learning Algorithms

Sheharyar Muhammad — 2024-06-01

The overall spread of the Coronavirus disease 2019 (COVID-19) irresistible sickness came about with a pandemic that has compromised a large number of lives. Infodemics is a well-known problem of interest. Nowadays, social media platforms are excellently representing the public sentiments and opinions about current events. Twitter is one of the most popular social media network that has captured the attention of researchers for studying public sentiments. Pandemic and Infodemics prediction on the basis of public sentiments expressed on Twitter has been an intriguing field of research. In this study we are focusing on people who have highest number of followers in Pakistan with most tweets related to Covid-19. The manually annotated dataset contains 2000 tweets of 1000 users for training and 380 tweets for test data from June to July 2020. For data processing we have manually labeled and added features to the dataset with the help of Senti Word Net. In the proposed model, KORONV is collecting data of tweets which will show the hashtags of COVID-19. Multiple machine learning algorithms are applied and the Long Short Term Memory (LSTM) gives the best accuracy of 98%. These techniques will be used to recognize patterns on the basis of existing algorithms, data sets and to develop adequate solution concepts that will be used for identification and classification between positive negative and neutral sentiment classification.

Skin Cancer Detection Using Deep Learning Algorithms

Saima Ali Batool — 2024-06-01

Skin disorders are very difficult to diagnose since many conditions have similar appearances, which makes it hard to tell them apart. While melanoma remains the most prevalent form of skin cancer, other illnesses are now known to be fatal. The most significant barrier to the development of automated dermatological systems is the scarcity of thorough and thorough data.This thesis introduces a robust deep learning architecture tailored for skin cancer diagnosis through the use of transfer learning on two types of convolutional neural networks (CNNs). These consist of a basic classifier and a two-tiered hierarchical classifier that makes use of two advanced CNNs. Differentiating between seven categories of nevi is the goal because precise diagnosis and treatment planning depend on it. The study's primary dataset is the HAM10000 dataset, a sizable collection of dermoscopic photos. Model performance is improved through the process of integrating multiple data sources. The outcomes unequivocally show how successful the DenseNet201 network is in this situation. This special combination reduces false positives while improving classification and Fmeasure accuracy. The test revealed that, surprisingly, the simple model performed better than the hierarchical two level model. The hierarchical approach is more straightforward than that, despite its attempt to provide classification at various levels. Specifically, the most successful level of binary classification is the first one, especially when it comes to differentiating lesions with and without nevus. This paper emphasizes the significance of applying deep learning methods particularly transfer learning to address the challenging skin cancer categorization issue. It is stressed how crucial data sets like HAM10000 are to the development of dermatological research. The outcomes validate the effectiveness of DenseNet201 in categorizing skin conditions and emphasize the necessity of refining the classification algorithm to produce more dependable, precise, and enhanced diagnoses, hence enhancing dermatological care.

Diagnosing Glaucoma Using Fundus Images

Mohabbat Ali — 2024-06-01

This paper presents a novel approach for the diagnosis of glaucoma, a leading cause of blindness, through the analysis of fundus images. Glaucoma is often marked by an insidious onset, with symptoms such as ocular redness, rainbow-colored halos, and gradual vision loss emerging as the disease progresses, ultimately leading to optic nerve damage. Given the subtlety of its early manifestations—earning it the mark "silent thief of sight"—early detection is critical. This research leverages advanced image processing techniques, including 2D Gabor filtering and Circular Hough transform, to enhance the features within retinal images that are indicative of glaucomatous changes. By integrating image segmentation through thresholding and match filtering, the proposed system effectively identifies the optic disc, a pivotal step in assessing the disease's presence. The methodology outlined herein demonstrates significant improvements in precision over existing diagnostic methods. This study lays the groundwork for future advancements, with the ultimate goal of augmenting the precision and ease of glaucoma detection for early intervention, thereby preserving the vision and quality of life for patients at risk.

Enhanced Skin Disease Diagnosis through Convolutional Neural Networks and Data Augmentation Techniques

Muddasar Abbas — 2024-06-01

Skin diseases are among the most common and widespread diseases affecting people around the world. Global warming and climate change are the two main factors leading to skin cancer. Skin diseases can be life-threatening if they are not detected and treated early. Clinical diagnosis of dermatological diseases is a relatively rare and expensive procedure, inaccessible to the general population. Advanced machine learning and image processing technologies enable feature extraction and disease identification. Deep learning is one of the emerging fields of machine learning that uses advanced intelligent techniques to classify images based on various features. In this research work multiple several deep learning models based on convolutional neural network (CNN) architecture such as sequential CNN model, DenseNet-121 and ResNet-50 are used for feature extraction. The HAM10000 dataset is used to evaluate the accuracy of the proposed model. There are 10015 images in the HAM10000 collection, which have been divided into seven different classes of skin diseases. The data set is divided into a training set and a test set, with ratios of 20% and 80%, respectively. Diagnosing skin diseases is a three-step process: first pre-processing the images of the dataset, then extracting features using CNN model in the second step and ends with classification of the skin disease category using various classifiers based on the features extracted in the third stage. Techniques for image augmentation were applied to lower the imbalance between different categories of skin diseases. The sequential CNN-based model with seven convolution layers achieves an accuracy of 98% and 99% for each of the seven categories of skin diseases in the Area Under the Curve (AUC). DenseNet-121 and ResNet-50 provide accuracy values of 89% and 84%, respectively. Various performance matrices are used to compare and evaluate the effectiveness of various CNN models on the provided dataset.

Navigating the Infodemic: The Impact of Social Media Rumors on Public Response to the COVID-19 Pandemic in Pakistan

Ramesha Rehman — 2024-06-01

Banknote Verification using Image Processing Techniques

Barrera Nazir — 2024-06-01

Automatic verification of the graphical data plays an increasingly essential role in the global financial system. Verification of banknote is a challenging problem for a human being to check the genuine currencies correctly. In dealings, generating counterfeit banknotes causes loss to the banks, individual, money exchange companies and many more. It has become tough for a human being to verify the currencies easily and appropriately. To avoid counterfeit banknotes, every nation-state comprises numerous kinds of security features such as watermark, flag, micro-printing and many more. Limited banknote verification models have been proposed in the past. Though, the earlier models suffer from a number of limitations which place strong obstacles to the real world banknote data sets. There is a dire need for a reliable technique to detect fake banknote. Based on this evaluation, the new framework is proposed to help the human being to discriminate between genuine and counterfeit banknotes. The proposed technique is based on the statistical features of a banknote such as color, shape, texture. The selective set of features is extracted with the help of Gray Level Co-occurrence Matrix (GLCM) and later on features are optimized using Principal Component Analysis (PCA). After extracting the set of features, three machine learning classifiers is applied to check the performance of banknote namely Decision Tree, Support Vector Machine (SVM) and K-Nearest Neighbor (KNN). The experiment results outperform the accuracy of proposed method.

Alleviating the Risk of COVID-19: A Social Healthcare Face Mask Detection System Based on Deep Learning Techniques

Moosa Khan — 2024-06-01

The purpose of a facemask detection-system is to ascertain in real-time whether or not an individual is wearing a face mask using an automated process. It commonly entails the utilization of Machine-Learning (ML), Deep-Learning (DL), and Computer-Vision (CV) approaches to examine images or video streams collected by cameras. The ability of the technology to distinguish between those wearing face masks and those who do not helps in the implementation of mask-wearing laws in a variety of settings, such as schools, hospitals, airports, and public transportation. One tactic being suggested to contain the spread is the wearing of face masks by persons in public areas. As a result, computerized face detection methods that are both successful and efficient are essential for such requirements. This article aims to design and implement an intelligent system that can detect faces that have been masked to mitigate the risk of COVID-19. MobileNetV2 makes integrating the system into devices with limited processing power easy. The photos are divided into two groups by this model: "with mask" and "without mask." During the model's development, it is trained and evaluated using a dataset of around 6,369 photos. The pre-trained MobileNetV2 model tha t we employed for this research achieved a 98.20% accuracy rate in terms of performance. Compared to VGG-16 and Inception-V3, the proposed system outperforms them in terms of its computational efficiency and accuracy. This work can be used as a digital verification tool in hospitals, colleges, banks, airports, and other public areas. This system can potentially improve public safety efforts and aid in preventing the spread of infectious diseases.

Beyond Classification: Exploring the Potential of NLP and Deep Learning for Real-Time Sentiment Analysis on Twitter

Waleed Ahmed — 2024-06-01

Twitter has evolved into a pervasive societal force, serving as a platform for diverse expressions, including official statements, thoughts, and opinions. This study delves into the multifaceted nature of Twitter, conducting a sentiment analysis on an extensive dataset comprising 1.6 million tweets and an additional 24,000 tweets. Leveraging advanced techniques in Natural Language Processing (NLP) and employing machine and deep learning algorithms, our focus lies in refining sentiment analysis methods to identify and mitigate terrorist-related content within social media posts. The study initiates with a comprehensive data preprocessing phase, involving part-of-speech tagging, sentiment score assignment via SentiWordNet, and neutralization of domain-specific words and negations. This meticulous methodology enhances the quality of the data for subsequent analysis. Weighted sentiment scores are then calculated, categorizing tweets into positive, negative, or neutral sentiment categories. To assess the effectiveness of our approach, various machine learning and deep learning algorithms are employed, including ensemble methods such as majority voting and stacking. Results reveal that Bidirectional Recurrent Neural Networks consistently outperform other models, achieving remarkable accuracy rates of 96% and 98% on two distinct datasets. Furthermore, the study explores diverse feature extraction techniques, shedding light on their impact on model performance. The findings of this research contribute to the ongoing discourse on sentiment analysis, particularly in the context of identifying and addressing potential threats within social media.

Identification of Scientific Researchers at the Early Stage of Field of Study Trends

Lubna Zafar — 2024-06-01

Classifying researchers at the emerging phase of a Field of Study (FoS) trend is of crucial. This process will reveal the early influential authors and guage the popularity of a particular FoS trend. Researchers might not only be active in emerging FoS trends relevant to their fields, but they might also find it highly helpful to be kept informed about the progresses of important new research areas. Companies and institutional funding agencies are also required to be frequently informed on changes to the scientific landscape, so that they can make initial choices about their important funds. The scientific community has produced numerous studies on the detection and analysis of FoS trends. These studies focus on multiple issues like, (i) birth and establishment of an FoS trend, (ii) number of publications and researchers in an FoS trend, (iii) communities of researchers being formed around an FoS trend, (vii) grouping of different FoS trends, etc. This study aims to identify authors active during the early stages of an FoS trend in the field of Computer Science. It utilizes scientific articles published between 1950 and 2018 within the Computer Science domain, sourced from the Microsoft Academic Graph (MAG) dataset. We have proposed an approach to detect influential researchers who were involved at the emerging stage of an FoS trend known as trend setters and the authors who followed it afterwards known as trend followers. The influential authors (trend setters) achieved high citation count and significance in a particular FoS. In our proposed approach, firstly, we have calculated the debut year of an FoS. Then, we have computed the FoS publication count, its author count and FoS trend by using Filed of Study Multigraph (FoM) with degree centrality measure. Afterwards, we applied Rogers' innovation diffusion theory for the detection of trend setters and followers. Lastly, we have compared our list of researchers (trend setters) with two existing lists of well-known Computer Science researchers. The following are the lists; (i) top 10 influential authors identified by [1] (ii) An existing list of Computer Science researchers with an H-index of 40 or higher (available at www.cs.ucla.edu/~palsberg/h-number.html) is utilized. The experimental results demonstrate that our proposed method successfully identifies many of the influential researchers featured on this list. In some instances, exact matches were found in relation to the FoS, confirming their status as trendsetters.

Diagnosis of Pulmonary Tuberculosis by Posterior-Anterior Lung X-Ray

Riasat Ali — 2024-06-01

Tuberculosis (TB) remains a pressing global health issue, with an estimated 10.6 million cases projected by 2021. In Pakistan, TB prevalence is notably high, comprising 61% of the WHO Eastern Mediterranean TB burden. TB, primarily caused by Mycobacterium bacteria, affects multiple organs, often presenting with subtle or asymptomatic symptoms. Despite the gravity of the disease, early detection methods are limited, typically relying on model lung segmentation techniques. This research aims to enhance TB detection using chest X-ray images through a novel, model-less segmentation approach. By extracting statistical, geometric, and Hog descriptor features from lung images, coupled with various classifiers such as Convolutional Neural Network (CNN), Support Vector Machine (SVM), and K-Nearest Neighbor (KNN), the study achieved promising results. The highest accuracy attained was 91.88% using self-extracted features and linear regression, while CNN demonstrated competitive performance with an accuracy of 89.58%. To bolster the findings, visualization techniques were employed, confirming CNN's superior ability to discern patterns from segmented lung areas, thereby contributing to higher detection accuracy. This innovative approach holds significant potential for expediting computer-assisted TB diagnosis, benefiting clinical practice and public health initiatives.

Towards deployment of IoT in ICT at Pakistan - Upcoming Challenges and Best Efforts Implementation

Mohammad Imran Mushtaque — 2024-06-01

Internet of Things (IoT) has become the new research area and being tested in closed environment to be deployed in real world. On the other hand, emergence of IoT in Information and Communication Technology (ICT) is an essential ingredient of technological development but even then it is not fully adopted especially in under developed countries. This research investigates the current status of IoT in ICT with respect to Pakistan and the challenges and efforts taken by the Pakistani Government and Institutions. This research reveals the current status and related issues of IoT in ICT so that swift measures may be taken to cope up with the situation in order to get enhance economic and technological advantage.

Blockchain-Based Decentralized Federated Learning for Privacy-Preserving Traffic Flow Prediction: A Case Study with PeMS-8 Data

Shaharyar Asad — 2024-06-01

A timely traffic flow information is essential for traffic management, traffic prediction has become a key component of intelligent transportation systems. However, current centralized machine learning-based traffic flow prediction algorithms need the collection of raw data for the train model, which poses significant privacy breach hazards. Federated learning, a recent innovation that effectively protects privacy by sharing model changes without transferring raw data, has been developed to solve these issues. The current federated learning frameworks, however, are built on a centralized model coordinator that continues to experience serious security issues, such as a single point of failure. In this paper, we proposed BDFL a block-chained decentralized federated-learning (DFL) architecture for traffic flow prediction using PeMS-8 data. The suggested technique provides decentralized model training on PeMS-8 while ensuring the privacy of the underlying data. The long short-term memory (LSTM) model, which is often employed for signal and time-series data, was used in this study. A total of 3035520 traffic locations were covered by the PeMS-8 dataset, which was obtained from the Caltrans Performance Measurement System (PeMS). These locations' data included timestep and junction details as well as traffic flow, occupancy, and speed. The total accuracy of our long-short-term memory (LSTM) model is 99.0, with a loss of 3.296.

Multifeature Analysis to Detect Cotton Leaf Curl Virus

Nazir Ahmad — 2024-06-01

Plant leaf diseases have devastating impacts on yield production, both in terms of quantity and quality. The cotton leaf curl virus (CLCuV) is one of the most destructive diseases that affect cotton crops worldwide. Disease detection based on symptoms is laborious and demands a great deal of experience and knowledge. The purpose of this research study is to design an automated system to detect CLCuV accurately. A dataset of healthy, mildly, and severely infected CLCuV is captured with a digital camera from cotton fields. An image enhancement tool is used to standardize the dataset for image analysis. Histogram, gray level co-occurrence matrix, and run length matrix features are extracted by the image analysis tool. Fisher, Probability of error plus average correlation and Mutual information feature optimization techniques are used to get the most optimal features to reduce computation costs. MultiClass, Bagging, Logistic Model Tree (LMT), and Radom Forest (RF) machine learning (ML) classifiers are deployed to observe the impact of CLCuV. True Positive (TP) rate, False Positive (FP) rate, Precision, Recall, F-measure, Mathews Correlation Coefficient (MCC), Receiver Operating Characteristic (ROC), and Precision Recall Curve (PRC) performance evaluation parameters are calculated to measure the effectiveness of ML classifiers. The RF classifier outperformed and demonstrated 87.542% accuracy, while other ML classifiers also achieved satisfactory results.

Integrating IoT and Machine Learning to Provide Intelligent Security in Smart Homes

Saira Batool — 2024-06-01

The technology is built entirely on the IoT and ML methods and is very desired in the sector of security. This method allows consumers to secure and regulate their homes. The system is used to track the treats in real time with response mechanism that detect with different sensor and send feed security information through internet. Smart home offer intelligent, real-time threat detection and response capabilities by utilizing decision tree, transfer learning logistic regression and SVM models. Smart home collects, preprocessed and trained data from many IoT sensors, such as cameras and motion detectors and then apply Ml models to detect the threads and perform the action according to requirements. A total accuracy of 0.992 indicated the model’s strong performance, confirming its suitability for practical use. Decision tree and transfer learning models perform higher Accuracy rates than SVM and logistic regression. These findings highlight the great potential of fusing IoT and machine learning technologies to produce flexible, effective, and scalable security solutions. They also offer a strong foundation for the implementation of dependable and high-performing machine learning models in real-world smart home security systems.

Use of Big Data in IoT-Enabled Robotics Manufacturing for Process Optimization

Farwa Abbas — 2024-06-01

Integrating big data analytics into IoT-based robotic manufacturing is essential to optimize processes and improve the efficiency of manufacturing environments. This study work describes the impact of big data analytics on IoT-based robotic manufacturing with a focus on process optimization and product improvement. To comprehensively evaluate the part of huge information analytics in process optimization, this consideration included the collection and investigation of different information parameters. These parameters are temperature, mugginess, control utilization, voltage, engine speed, torque, weight, vibration, stack capacity, operational productivity, generation rate, mistake code, and communication status. The information collection preparation was conducted utilizing IoT sensors conveyed over the fabricating office, guaranteeing the capture of real-time information for an investigation. In the classification of production conditions, three classifications were used based on the collected data - bagging, SVC, and decision tree. Each classifier has good advantages in analyzing complex data sets and identifying patterns that aid in informed decision-making and process optimization. In the context of a study on the use of big data analytics in IoT-based robotic manufacturing, the decision tree classifier shows a high accuracy of 97%. The bagging classifier achieved 94.39% accuracy, while the Support Vector Classifier (SVC) achieved 96% accuracy. This research explores machine learning analysis methods and address ethical issues to maximize the benefits of big data analysis in IoT-based robotic manufacturing.

AI and Sensing -Enhanced Irrigation through Cable Rail for Drought and fros Prone Regions in the Face of Climate Change

Attique Ur Rehman — 2024-06-01

Potato and lychee are major crops in Pakistan. Lychee, are called queen of fruit in emerging of fruit crop, only Lychee fruit cover 1.23kha area of Pakistan. Due to hot and dry weather, the skin of Lychee cracks. Potato another crop of Pakistan those cultivate in huge area almost 160kha.potatoa crop also effective with harsh weather like as frost burning. In generally, farmer prevents this problem by burning fire and splitting water for potato crops and for lychee sprayer water in the summer season but in this scenario, timely activation of phenomena is very important otherwise ruin the crop. In the agricultural practices. Here we explore the role of IoT, sensing technology and A.I strategies in the agriculture industry that can help to improve farming practices. Agriculture faces many provocations like food security, production, and water conservation due to sudden climatic changes, high temperatures, and uneven distribution of resources. To manage all the event. We address the Smart Rail Irrigation System (SRIS) that deal with the pressing issue and give transformative result in the yield of crops. The objective behind using this technology is to monitor, manage resources, and resolve harsh weather relative problems. Smart solution for irrigation, disease prevention sprays and proper distribution of fertilizers will also help in the conservation of water. This automated system in agriculture is beneficial for field coverage, efficiency, and highest yield production, improving food quality, and conserving water. SRIS system is especially designed for two types of crops Lychee and potato.

Benchmarking of an Enhanced Grasshopper for Feature Map Optimization of 3D and Depth Map Hand Gestures

Fawad Salam Khan — 2024-06-01

The Enhanced Grasshopper Optimizer (EGO) for the feature map optimization of 3D and depth map hand gestures is the objective of this paper's benchmarking experiment. Using a dataset of 3D and depth map hand gestures, the effectiveness of the EGO algorithm is examined and contrasted to alternative optimizers. The optimized feature map is tested using the Rosenbrock benchmark test function with EGO and SGD, the findings demonstrate that the EGO algorithm performs better than the alternative techniques in terms of precision and computational time. The execution time of EGO is also benchmarked in this study with the different numbers of input features and shows dominance in performing feature selection for 3D hand gesture detection and classification.

An Analysis of Supervised Machine Learning Techniques for Churn Forecasting and Component Identification in the Telecom Sector

Hira Farman — 2024-06-01

The term "business intelligence" (BI) refers to a broad range of tools and software intended to collect, process, and evaluate data so that business users may decide on the best course of action. Getting the right information to the right decision-makers at the right time is the main goal of business intelligence (BI). The telecom business creates massive amounts of data every day because of its big clientele. Decision-makers and business professionals emphasized that maintaining existing clientele is less expensive than recruiting new ones. In addition to detecting patterns of behavior from the data on existing attrition clients, business analysts and customer relationship management (CRM) analysts need to understand the reasons for customer attrition. This paper focuses on customer churn, a critical metric that represents the percentage of customers ending their relationship with a company over a specific period. Using detailed datasets and advanced data analysis and machine learning techniques, the key churn predictors were found in this study, along with practical recommendations on how to improve customer retention. In order to create predictive customer churn models, several important machine learning algorithms have been surveyed and compared in this study. This study looks at more than just churn and non-churn classification; it also evaluates the accuracy of different data mining techniques. Uses a variety of performance indicators and confusion matrices to assess the effectiveness of three classification models: Random Forest (RF), Decision Tree (DT), and Logistic Regression (LR). With the best AUC (0.985), F1 score (0.934), Precision (0.935), and MCC (0.830), the Random Forest model outperformed the others, demonstrating a strong balance between recall and precision. The Decision Tree model also performed well, with notable accuracy). Logistic Regression, while effective, showed comparatively lower metrics, with an AUC of 0.848 and an F1 score of 0.800. The confusion matrices further validated these results, highlighting the Random Forest model's robustness and superior classification capabilities. The findings show that with the RF algorithm, our suggested churn prediction model generated superior churn categorization. Furthermore, this research delves into the fundamentals of BI and presents optimization strategies crucial for making dynamic, optimal decisions in today’s corporate landscape.

Analysis and Clustering of Pakistani Music by Lyrics: A Study of CokeStudio Pakistan

Zia Ur Rahman — 2024-06-01

This research explores the application of unsupervised learning techniques to categorize and understand the lyrical content of CokeStudio songs. In a world where music transcends cultural boundaries, this study delves into the rich linguistic tapestry of lyrics, unraveling emotions, themes, and cultural nuances. We begin by employing Natural Language Processing (NLP) and analysis techniques to uncover the emotional underpinnings of these lyrical compositions. This emotional layering becomes the foundation for the subsequent clustering process. Multiple unsupervised learning algorithms, including K-Means, Hierarchical Clustering, and DBSCAN, are employed to categorize songs into thematic clusters. The quality of these clusters is assessed using the silhouette score, with the optimal number of clusters determined as 5, achieving a score of 0.41641. Furthermore, we develop a robust classification model utilizing machine learning algorithms such as Logistic Regression, Decision Trees, Random Forest, Gradient Boosting, and Multinomial Naive Bayes for evaluation of our clustering. This model assigns CokeStudio songs to thematic clusters based on the results of topic modeling, enhancing our understanding of the cultural and emotional dimensions of these compositions. Logistic Regression, with SMOTE applied to NMF values, emerges as the best-performing model, achieving an impressive testing score of 89.47%. The research findings not only illuminate the intricate emotions and narratives woven into CokeStudio songs but also emphasize the practical application of machine learning in music analysis. By identifying and classifying thematic clusters within song lyrics, this study enriches our comprehension of cultural expressions through music and opens avenues for personalized music recommendations.

Optimizing Pneumonia Diagnosis during COVID-19: Utilizing Random Forest for Accurate Classification and Effective Public Health Interventions

Fazal Malik — 2024-06-01

In the quest for precise diagnosis and classification of pneumonia, particularly intensified by the Coronavirus 2019 (COVID-19) pandemic, this research work presents an optimized Random Forest algorithm based mechanism specifically tailored for COVID-19 pneumonia classification. The research methodology encompasses four critical phases: data acquisition from a current COVID-19 X-ray image dataset on GitHub, data processing and analysis using histograms and scatter plots, application of supervised learning with Random Forest enhanced by data augmentation techniques, and performance evaluation through comparative analysis with existing methods. Our proposed model achieved an accuracy score of 82.29% on average, demonstrating significant precision and recall capabilities. Results indicate that the Random Forest model outperforms current methodologies, providing a robust framework for future pneumonia classification research. This study underscores the potential for improved diagnostic accuracy and patient care, highlighting the model's utility in supporting public health interventions and optimizing resource allocation in the context of COVID-19 pneumonia.

Role of Kubernetes in DevOps Technology for the Effective Software Product Management

Hira Haroon Khan — 2024-06-01

DevOps is a set of methodologies and cultural values that automate and combine software development (Dev) and IT operations (Ops) to create the best application. DevOps, based on agile concepts, applies incremental creation and continuous input across the product lifecycle. DevOps reduces friction between operations and development to improve software development and delivery. These rules govern software efficiency, structure, naming, and documentation. Jenkins automates code development, verification, and distribution, enhancing CI/CD efficiency and dependability. Kubernetes streamlines container-based application administration, enabling controlled scalability, growth, and fast service execution. These technologies improve DevOps processes, ensuring on-time delivery of items. DevOps release management includes thorough planning, infrastructure, versioning, restoration plans, continuous tracking, automated testing, reflection alerts, interaction, and recordkeeping. Well-controlled processes ensure reliable application deployments by resolving issues quickly, simplifying setup, and reducing user effort. Kubernetes' Permanent Amounts, flexible supply, and capacity divisions help manage, customize, and extend storage. It streamlines deployment and maintenance, organizes containers into Pods, and provides a flexible and robust environment due to its formal approach. Kubernetes manages all program storage, simplifying application deployment and management. Kubernetes has powerful networking, cloud support, recovery, load balancing, simple scalability, strong monitoring, and backup operations.

Intellectual Gesticulation Identification Assembly

Zainab Zafar — 2024-06-01

Public sign language recognition is an important step for a comminute gap between people, physically chal-lenged due to lack of hearing and speaking, with people who can easily convey their messages, using a sign lan-guage translator we convert given gestures to textual form in the form of alphabets/cat-digits. Hence making it simpler of recognizing the speech textual form and also how the gestures they passed on. Data acquisition We have collected a dataset of 44 gestures (which include all the alphabets and digits). In this paper, we present an-ticipated approach to detect the way of Intelligent hand gesture recognition system enabled by CNN. Things to do, We first preprocess our input image after then we have to remove photo noise from the image. Then apply the threshold to straight photos. Region filling: used to fill in holes in the object of interest. This results in a model with CNN keras using TensorFlow as backend for the trained data. Classify the training data. Data tests are per-formed by the keras model. Once the testing has done next feature is gesture recognition as the user pass the ges-ture and in result window displays in text format of a gesture and in speech form as well.

Arrhythmia Classification and Analysis on ECG Using Convolutional Networks and Two-fold Focal Loss

Mumtaz Ali — 2024-06-01

Throughout recorded history, cardiovascular diseases have posed a persistent threat, claiming numerous lives. Effective and timely testing is pivotal in preventing fatalities. Among the available testing options, the Electrocardiogram (ECG) stands out as both practical and cost-effective, capable of diagnosing various abnormalities. Recently, there has been a notable emphasis on accurately classifying heartbeats. Traditionally, heartbeat analysis has been approached through manual or automated methods. Manual analysis involves cardiologists, while automated analysis relies on computational algorithms. Automated techniques have gained significant popularity in recent years and have achieved considerable success. However, despite this progress, there is still a need for further improvement to achieve deployable accuracy. Many current studies utilize deep learning models in a transfer learning approach for heartbeat classification. While transfer learning offers advantages, it also presents disadvantages such as domain mismatch, task-specific features, interpretability concerns, model bias, and generalization issues. Therefore, in this study, instead of employing transfer learning, a deep convolutional neural network combined with twofold focal loss is utilized for heartbeat classification. The proposed approach has demonstrated the ability to accurately classify five distinct arrhythmias according to the AAMI EC57 standard. Testing was conducted using the MIT-BIH and PTB Diagnostics datasets from PhysionNet. The results indicate that the proposed method achieves an average accuracy of 99.8% in classifying arrhythmias.

A Comprehensive Review on DDoS Attack in Software-Defined Network (SDN): Problems and Possible Solutions

Fateh Ahmed — 2024-06-01

This paper provides key insights into the classification of distributed denial of service (DDoS) attacks and defensive techniques to protect software-defined net-works (SDN) in these attacks. The networking industry is evolving due to a revo-lutionary paradigm known as software-defined networking. It is the type of net-work where data and control planes have been decoupled to minimize errors and enable efficient use of network resources. DDoS attacks have proved to be a major threat to any business such as small- and large-scale enterprises. These attacks have the potential to destroy businesses in a few hours. Even giants like Amazon have reported that it has thwarted one of the biggest DDoS attacks. Attackers tar-get various SDN-based networks to cause huge losses to entrepreneurs and busi-nesses. The most worrisome part is that they are still taking place across the globe in no time. Unlike a simple denial of service attack, many nodes initiate an attack on the network or server in a distributed environment. Attacks are seriously damaging the CIA triad: confidentiality, integrity, and availability. Moreover, the performance and security metrics of infrastructure are also affected. As things stand, we need to realize that we can avoid DDoS to some extent but not com-pletely.

Object Identification for Autonomous Vehicles using Machine Learning

Hafiz M. Mubeen Ahmed — 2024-06-01

Autonomous vehicles (AVs) hold immense promise in reshaping transportation by enhancing safety and efficiency. A critical challenge lies in accurately identifying objects at long ranges, particularly in adverse conditions. This study explores the application of machine learning algorithms for the long-range object identification in AVs. Methodologically, a diverse dataset encompassing real-world data from multiple sensors is curated and preprocessed. Various machine learning models, including convolutional neural networks (CNNs), recurrent neural networks (RNNs), and deep reinforcement learning (DRL), are trained and evaluated using this dataset, with metrics such as accuracy, precision, recall, and F1 score employed for assessment. Results indicate promising performance, with sensor fusion techniques augmenting accuracy and reliability. Ethical considerations are addressed, emphasizing safety and bias mitigation. Limitations of current models in terms of robustness and generalization are analyzed, alongside proposals for enhancement. Findings underscore the significance of sensor fusion, model validation, and data diversity in ensuring AV safety and reliability. In conclusion, this research advances the field of AV perception systems, laying a foundation for safer and more efficient autonomous transportation.

Techniques for Authentication and Defense Strategies to Mitigate IoT Security Risks

Ayesha Munir — 2024-06-01

Internet of Things (IoT) has rapidly evolved into a transformative technology, impacting various facts of daily life and industry. However, its widespread adoption has been impeded by significant challenges, particularly in the realms of security, privacy, interoperability, scalability, and Quality of Service (QoS). These challenges represent critical obstacles that must be addressed to ensure the reliability and efficacy of IoT systems. It addresses security vulnerabilities through multi-layered measures including encryption, authentication mechanisms, anomaly detection, and secure firmware updates, while also prioritizing privacy preservation through privacy-by-design principles and data anonymization techniques. Additionally, methodology advocates for interoperability frameworks, scalability strategies encompassing cloud-based architectures and edge computing paradigms, and Quality of Service enhancements through performance monitoring and adaptive resource allocation. By implementing these methodologies, we aim to overcome the fundamental challenges facing IoT deployment and pave the way for a more secure, interoperable, scalable, and reliable IoT ecosystem, underscoring the importance of comprehensive solutions to unlock the full potential of the Internet of Things.

Empowering Creativity: Exploring Makerspace Realities and Challenges in Public Sector University Libraries of Khyber Pakhtunkhwa, Pakistan

Afsheen — 2024-06-01

Purpose-The purpose of this study is to investigate the current status and challenges faced by makerspaces in public sector university libraries of Khyber Pakhtunkhwa, Pakistan, aiming to provide insights for enhancing their effectiveness in fostering creativity and innovation among students and researchers.

Methodology/ Research approach- This study utilized a quantitative survey with a structured questionnaire, involving 80 librarians from 34 public sector universities in Khyber Pakhtunkhwa. The questionnaire, based on current literature, was distributed via email, Google Docs, and social media platforms for wide participation. Total 69 responses were received from the university librarians. Data analysis was conducted using SPSS software, focusing on descriptive statistics and Likert scale responses.

Findings/Outcomes of the study- The analysis revealed a moderate perspective regarding the availability of tools and engagement in diverse projects, reflecting the current status. Challenges encompassed staffing expertise, technology infrastructure, funding allocation, awareness promotion, institutional support, and curriculum integration. Proposed improvement strategies focused on staff training, securing diverse funding, integrating curricula, fostering external collaborations, enhancing community engagement, and promoting interdisciplinary initiatives. These findings highlight the necessity for holistic strategies to bolster makerspace effectiveness and sustainability within university libraries.

Practical implications– This research provides valuable insights for policymakers, senior administrators, and library managers in establishing makerspaces in prestigious academic institutions. Key recommendations include enhancing staff training, diversifying funding sources, integrating makerspaces into curricula, promoting collaborations, fostering community engagement, prioritizing funding, and improving institutional support. These strategies are crucial for optimizing makerspace effectiveness and sustainability in university libraries, making our findings essential for decision-makers shaping education and innovation in these institutions.

Innovative Machine Learning Techniques for Malware Detection

Aqsa Ijaz — 2024-06-01

Malware hazards are becoming more perplexing with time, new types of malware are entering cyberspace and triggering millions of devices day by day. People could not restrain in this century to refrain from not using smart devices, and adopting technology, as this world is shifting into a smart world, and due to the COVID19 wave, more numbers of devices and systems were being adopted by the people. In viewing the need of the society and to save the cyber world we have to step into this war against cybercrimes and play our role to save this world by making such models that are efficient and effective against malware. Therefore, accordingly, machine learning techniques have become the main point for cybersecurity as they are most suitable for handling modern malware attacks. Moreover, machine algorithms can generalize and distinguish cyber threats to a great extent. We applied an ensemble model in which we have used different machine learning algorithms such as KNN, SVM, and LR, as first stage classifiers and voting classifiers as meta-learner classifiers to identify the complex and modern malware. We have applied hard voting in our ensemble model. We also discuss and evaluate the performance of every algorithm applied in the model. KNN shows the best results overall. The ensemble model provides us the best result than any individual used model. The output of testing proves that our proposed method is highly efficient and adaptive and gives better results than many other present techniques. We gain 99.7 % accuracy with F-score 99%. The running time of the model is also less. So this proposed detecting malware model could be easily implemented in smart IoT devices as well.

Lysine Acetylation Site Prediction in Prokaryotes: A Deep Learning Approach

Hassan Kaleem — 2024-06-01

Post-Translational Modification (PTM) of proteins plays a vital role in both disease and normal states. Protein acetylation is an important PTM in eukaryotes as it greatly changes the properties of a protein including hydrophobicity and solubility. Therefore, in both metabolism and regulatory processes, acetylation and other PTMs perform a critical role. By Investigating and accurately spotting lysine acetylation sites can stop or alter faulty modifications that were previously supposed to occur. This can help in changing the course of microbiological diseases like Bacteremia, UTI's, meningitis and others. Several models have been developed to identify lysine acetylation (Kace) sites with appreciable performances. This manuscript presents an improved approach to identify lysine acetylation (Kace) sites which achieves 0.951, 0.891, 0.813, 0.969, 0.946, and 1.0 MCC for B. subtilis, C. glutamicum, E. coli, G. kaustophilus, M. tuberculosis and S. typhimurium species respectively. Machine Learning algorithms require feature extraction from protein sequences, which is a complex and time taking process. This study has introduced an approach to identify kace sites using a deep learning-based model. The proposed approach significantly outperforms the existing approaches. The experimental results on the benchmark and independent datasets achieve significantly higher accuracy, very close to the actual labels. The source code accurate prokaryotic-lysine-acetylation-site-prediction for the proposed approach is made publicly available online for validation purposes.

Analyzing Machine Learning Models for Forecasting Precipitation in Australia

Hira Farman — 2024-06-01

In the 21st century, predicting when it will rain is an intriguing but challenging task. Climate and precipitation representations are frequently extremely complicated, non-linear, and inconsistent, need highly skilled, specialized mathematical modeling and training. The rise in rainfall-related flood tragedies in recent decades has made weather forecasting an increasingly important area of study. Most of the time, the researcher tried to find a linear relationship between the necessary data and the meteorological data that was already accessible. This work uses conventional machine learning algorithms to give a thorough analysis and prediction model for rainfall in Australia. Enhancing the precision and dependability of rainfall forecasts is the aim of this research. The dataset for the study contains historical meteorological data, such as temperature, humidity, wind speed, and air pressure, from multiple locations of Australia. Using classic machine learning techniques like Random Forest (RF) and Naive Bays closest neighbors, baseline models are created. Model evaluation is a meticulous procedure that contrasts the accuracy, precision, and memory of models. The primary meteorological factors that influence the variability of rainfall are identified using the feature importance analysis. The interpretability of the models is also investigated in the study in order to offer insightful information about the decision-making procedures. The dataset includes 14, 5460 size, 23 features detailed city-specific monthly averages for Australia from 2008 to 2017(10 years). An effective rainfall forecasting was produced by integration of a number of machines learning techniques, including Random Forest model (RF), K nearest Neighbor (KNN), Decision Tree (DT), Naïve Bayes (NB), and Logistic Regression (LR). This research intends to mitigate the high risks of floods induced by natural disasters by utilizing state-of-the-art models. The results show that random forests have high accuracy (0.859) for predicting rainfall.

Architectural Formation with Deep Learning and Algorithmic Bindings for Cross-Domain Information Retrieval

Khawaja Tehseen Ahmed — 2024-06-01

Efficient strategies for index search are crucial elements involved in categorizing and retrieving simple as well as complex image collections and libraries. In this paper, new algorithm is presented aimed at refining the selection of images to be clustered and more accurate identification of ROIs in many clustered objects. The relations to other features are also expected to be provided, including the RGB image features and the other feature sets obtained with the use of Convolutional Neural Networks (CNNs) for achieving the scale invariance. Despite, GoogleNet and AlexNet and ResNet exist this algorithm has the deep feature and spatial data point of view for improving the image classification. Feature coefficient computation further enables the application of norms L1 and L2 on over the images of RGB. The ‘Scale invariance’ encompasses predicting the scaling of keypoints, computation of coefficients between two successive octaves along with expressions of virtual intra octave expressions. In the process of maxima selection, interpolation, non-maxima suppression, and cumulative thresholding the algorithm applies ROI detection. The presented multimodal approach significantly enhances the identification of objects particularly in a setting as depicted in this paper with high density of other similar objects. The color feature sets and CNN feature sets that are integrated in constructing the Bag-of-Words (BoW) model improve image indexation and image search. From the quantitative analysis, there is promising average precision (AP) and average recall (AR) when the presented algorithm is tested using data from Corel-10K, Tropical Fruits and Cifar-10 datasets.

Air Quality and Carbon Monoxide Monitoring Using IOT-based System

Muhammad Kamran — 2024-06-01

The humans utilize around 95% of indoor air and air quality monitoring is essential for maintaining healthy environment. The Internet of Things (IoT) contributes to facilitate smart and safe lifestyle. The pollution monitoring is essential to avoid exposure to dust particles and harmful gases and minimize its impact on the surrounding environment. The paper presents an effective Air Quality Monitoring System (AQMS) to evaluate temperature, humidity and concentration of gasses like Carbon Dioxide (CO2), Carbon Monoxide (CO) respectively. It consists of Microcontroller (NodeMCU ESP32), Temperature and humidity sensor (DHT-11), Carbon Dioxide sensor (MQ-135), and Carbon Monoxide (MQ-7). Sensors fetch the environmental data and transmit to the Microcontroller (NodeMCU). The microcontroller generates alert for breach in normal concentration of environmental parameters. Moreover, the system is capable to open doors and windows through the servo motor. The model is capable to send information through a wireless link to blynk application. The Test-Driven Development Methodology for Internet of Things-based Systems (TDDMIoTS) and the technologies used to automate the creation of Internet of Things systems have simplified the monitoring mechanism. This system is easy to install in houses, industries and vehicles for protection of humans working as well as living in indoor and outdoor places.

A Deep Learning Framework Based on GCN Model for Android Malware Detection

Syeda Huma Zaidi — 2024-06-01

Nowadays, Android malwares are increasingly significantly producing major security issues. The complexity and increase of malware threats have made automated malware detection research an important component of network security. Traditional malware detection methods include manual examination of every malware file present in the application, which consumes a significant number of human resources (on the basis of both storage and time). Additionally, malware developers have created methods like code obfuscation to get beyond antivirus companies' conventional signature-based detection methods. Deep learning (DL) approaches for malware detection are now being used to resolve this issue. In this study, Performance comparisons are made amongst GCN (Graph Convolutional Network) models for Android malware detection. Using graph-based representations of malware of the Android DEX file, a GCN-based model is suggested to detect Android malware. GCN extracts the necessary features from the images of malware. The static approach is used to extract the essential features. Then, these features train GCN to detect malware. We presented a GCNs latest version for modeling more advanced graphical semantics. It automatically discovers and understands semantic and ordered patterns based on the previous stage's vectors, without requiring additional sophisticated or expert features. The proposed method outperformed the compared models in every performance metric, achieving an accuracy of 99.69% compared to other approaches.

Automated Brain Tumor Detection via Transfer Learning Techniques

Moeez Bin Nadeem — 2024-06-01

Brain tumors disrupt the regular operation of the brain, and if left untreated, these malignant cells can impact the adjacent tissues, blood vessels, and nerves. Moreover, it affects a large number of individuals worldwide and can result in substantial damage. Thus, it is crucial to understand that brain tumors are a severe medical disease that demands appropriate medical attention. Tumors are the primary cause of a significant number of deaths in modern times. They damage the brain, leading to severe mental as well as physical problems. Detecting brain tumors manually is a difficult task because of variations in their appearance, such as differences in shape, size, and nucleus. Consequently, there is a need for an automated approach to detect brain tumors at an early stage. This paper presents a study on detecting brain tumors utilizing a “Convolutional Neural Network (CNN)” with the “Adaptive Moment Estimation (ADAM): optimization algorithm. Using transfer learning, the researchers built a base model in CNN and combined it with “RESNET-152”, MobileNet, and Densenet-121. The classification of brain tumors as either tumors or non-tumors was performed and evaluated on a public Kaggle brain-tumor dataset. The results showed that the proposed model achieved 98.7% accuracy and 99.8% AUC for Resnet, 96.5% accuracy and 98.6% AUC for Dense Net, and 87.2% accuracy and 98.7% AUC for MobileNet, respectively. The data indicate that using the RESNET-152 model produced better results than other baseline approaches. This suggests applying the model to additional disorders and its clinical utility in routine practice.

An Indexing-Based Architecture for Fast Data Retrieval in Smart Cities

Muhammad Bilal Aslam — 2024-06-01

A Smart City leverages Information and Communication Technologies (ICT) to enhance the quality of life for its residents by improving operational efficiency and providing reliable services. The primary objective of a Smart City is to use technology for optimization and decision-making. This involves deploying a network of sensors connected to the internet to collect real-time data from the environment, which is then shared across various systems to make intelligent decisions and improve city operations. Smart City applications encompass smart homes, transportation systems, traffic management, power consumption management, water supply networks, and other community services. The resulting vast amounts of data, known as Big Data, present significant challenges in terms of acquisition, management, and real-time analysis. Despite various proposed solutions by researchers, efficient data handling remains a critical issue. In our study, we propose an innovative architecture based on indexing to enhance data analytics for Smart City applications. This architecture covers data acquisition, storage, and retrieval. We validate our proposed architecture through the analysis of diverse datasets related to Smart Cities, demonstrating its effectiveness in managing and processing large-scale data efficiently.

Optimizing Road Extraction with Residual U-Net: Enhanced Training

Shah Zaib — 2024-06-01

Computer vision and remote sensing depend heavily on extracting roads from satellite or aerial photos. This study presents a novel approach to road extraction employing a Residual U-Net architecture with integrated data augmentation techniques. The proposed method utilizes a deep learning model with residual blocks for improved feature extraction and semantic segmentation. The dataset is preprocessed, and data augmentation is applied during training to enhance model robustness. The augmentation includes random, non-critical actions and horizontal flips. The Residual U-Net architecture consists of an encoder-decoder structure with skip connections, facilitating the learning of intricate spatial dependencies. The model is trained to optimize road segmentation using a customized loss function based on the Dice Coefficient. Additionally, the code incorporates batch normalization and activation functions for improved convergence and generalization. The experimental findings show how effective the suggested strategy is for road excavation jobs. Training and validation sets are generated using a custom data generator class. The model is trained over several epochs, and its performance is evaluated on a validation set. Ground truth versus predicted value visualizations showcases the model's ability to delineate road networks accurately. This study contributes to road extraction by introducing a Residual U-Net architecture with data augmentation, providing a robust and accurate solution for road segmentation in satellite imagery.

A Comprehensive Survey on Security Threats and Challenges in Cloud Computing Models (SaaS, PaaS and IaaS)

Ezzah Fatima — 2024-06-01

Cloud computing has fundamentally transformed data access, storage, and processing for individuals and businesses alike, offering unparalleled scalability, flexibility, and cost-effectiveness through Software as a Service (SaaS), Platform as a Service (PaaS), and Infrastructure as a Service (IaaS) models. However, as with any revolutionary paradigm shift, cloud computing is not without its challenges and concerns, cybersecurity being chief among them. This research article examines security challenges and attacks in each cloud computing model, with SaaS SQL injection and deceitful QR code attacks highlighted, PaaS facing vulnerabilities like unauthorized access addressed through the Multi-Perspective PaaS Security Model, and IaaS confronting issues such as data breaches by emphasizing shared responsibility. To address these challenges and concerns, this study proposes and explores potential solutions, such as integrating machine learning and encryption techniques to mitigate vulnerabilities and enhance the security posture of cloud platforms. By discussing relevant literature and conducting comparative analysis of existing works, this study also identifies research gaps and trends that need to be addressed in the future, including emerging threats and standardization of cloud security measures that would contribute to the establishment of industry-wide standards and effective countermeasures in cloud computing. The findings of this study underscore the critical importance of collaborative efforts between users and Cloud Service Providers (CSPs) in addressing cybersecurity concerns, as well as the need to enhance user awareness and adopt robust longitudinal studies to mitigate future security risks. Ultimately, this research aims to provide a comprehensive understanding of evolving security landscapes in cloud computing, and contribute to the establishment of industry-wide standards and effective countermeasures in cloud computing.