Journal of Computing & Biomedical Informatics

Modified U-Net Model for Segmentation and Classification of Liver Cancer Using CT Images

Zunaira Naaqvi — 2024-03-01

Liver cancer is becoming more common worldwide, and precise and accurate tumor segmentation is required for early detection. However, segmenting tumors is extremely difficult due to their hazy borders, variability in appearance, sizes, and varying densities of liver tumors. In the domain of medical images, a multimedia-based system is an ultimate requirement. It is a primary need in the healthcare industry that is also necessary for patients and doctors to achieve quick and efficient results. Deep learning-based approaches are currently being used to improve performance in various domains. Preprocessing, augmentation, segmentation, and classification are the four stages of the proposed framework. Because deep learning methods perform well on large datasets, the effects are evaluated of data augmentation synthetically and then used the five times transformation technique to increase the number of training samples. This paper proposes a deep-learning method for segmenting liver tumors based on a modified U-Net model called the AU-Net model. This framework employs an AlexNet CNN-based architecture to classify liver tumors. As a result, the 3D-ircadb01 and LiTs datasets were used, which are freely available. On the 3D-ircadb01 and LiTs datasets, the proposed architecture gained accuracies of 97.6% and 98.45% for liver classification. The proposed architecture consistently produces the best and most accurate results compared to other state-of-the-art methods.

Pakistan’s Political Sentiments Analysis based on Twitter Using Machine Learning

Arfan Ali Nagra — 2024-03-01

Pakistan’s Politics has always been a hot topic. Politicians wastes no time to express their sentiments and narratives through social media without knowing the whole situation. These social media posts and tweets create a situation which can lead to anarchy in the society. It is a need of hour to figure out which politician creates the more dramatic situation. It can be done through sentimental analysis of their social media account. Sentiment analysis aims to extract sentiments, opinions, and emotions from social media sites like Twitter. The conventional technique of sentiment analysis is concerned with textual data in which users post updates relating to various themes. This manuscript examines five Pakistani politicians consisting of Ex-Prime Minister Imran Khan, Vice President PML-N Marium Nawaz, Chairman PPP Bilawal Bhutto, Spoke-person PTI Shebaz Gill, and Information Minister PML-N Mariem Aurangzeb. The focus is to analyze how they have utilized Twitter to interact with their followers and, in doing so, influence the political process. Data are collected from Twitter accounts belonging to various Pakistani politicians. A comprehensive framework of pre-processing procedures for making tweets more manageable is presented. The main goal is to make sure that people get knowledge about the better direction for the society. In political sentiment analysis, each politician's tweets are classified into positive, negative, and neutral sentiments to provide a unique perspective on how hate speech is used by politicians. This is accomplished with a machine learning classifier i.e Support Vector Machine, Random Forest, Decision Tree and Logistic Regression. After comparative study of these classifiers, Random Forest achieved the highest accuracy 86% among all. Such classifier will aid organizations, political parties, analysts, and others in assessing public sentiments regarding them. As a result, the most negative tweets are used by Imran Khan.

Citrus Diseases Detection using Deep Learning

Nouman Butt — 2024-03-01

Pakistan is a major contributor to citrus fruit production, accounting for 30% of the total fruit output, with citrus cultivation spread across all four provinces, particularly Punjab. Citrus is vital to domestic and international markets and is distributed through various value chains. However, like many fruits, citrus is susceptible to diseases such as canker, citrus scab, and black spot, impacting fruit quality and quantity. Manual disease diagnosis in citrus fruits is time-consuming, error-prone, lacks standardization, and incurs high costs, requiring expert intervention. Accurate diagnosis and treatment are imperative for safeguarding citrus crops. The implementation of automated systems provides efficient, consistent, and cost-effective solutions, mitigating the challenges associated with manual diagnosis and contributing to sustainable citrus farming practices. This paper introduces an automated system employing Deep Learning and optimal feature selection for classifying citrus diseases. The process begins with data augmentation, enhancing the training dataset by creating additional images from existing samples. Two pre-existing models, DenseNet-201 and AlexNet, undergo adaptation and retraining utilizing an augmented dataset via transfer learning techniques. The experiment is carried out on the leaves dataset.Attaining the highest accuracy of 99.6%. The suggested framework is examined at every phase and compared to modern methods approaches, affirming its superior performance.

A Comprehensive Analysis of Machine Learning and Deep Learning Approaches for Road Accident Prediction

Hamid Ghous — 2024-03-01

Road accidents pose a substantial concern for all individuals. On a daily basis, a substantial number of precious lives are tragically lost due to vehicular collisions. The research conducted on road accident detection and prevention involves the utilization of various datasets to predict potential scenarios that may result in road accidents. Nevertheless, a significant obstacle that arises during the development of computer-vision models aimed at identifying traffic attributes in road accidents is the scarcity of available datasets. The dataset lacks diverse factors contributing to road accidents, which could lead to leveraging for training deep-learning models on a broad scale. This work seeks to give an in-depth examination of road accident databases, implementing machine-learning techniques, and use of Deep Learning (DL) algorithms on road accident datasets. This research examines the methods of Machine Learning (ML) and DL algorithms that are utilized in the process of creating road accident projections, as well as their relevance to the data sets that are being taken into consideration.

Skin Lesion Detection and Classification

Maaz Ul Amin — 2024-03-01

Skin cancer has an overall mortality rate of 0.6 to 0.7% and accounts for 5.8% of cases worldwide (1.6% for melanoma). Globocan 2018 is challenged by Pakistan's Punjab Cancer Registry (PCR), which records an increased incidence. According to PCR, skin cancer is eighth and ninth most common in both males and females in 2017. It ranks in the top eight for Karachi according to the Karachi Cancer Registry (KCR). According to Dow University of Health Sciences (DUHS), non-melanoma skin cancer is the sixth most common type of the disease, which indicates a notable increase in Karachi. This study identifies key contributors among universities, research institutions, and cities and objectively assesses Pakistan's skin cancer research, examining output, types, and focuses. Additionally, it intends to identify the main institutions and cities that have made significant contributions to this field of research. The research presents a sophisticated computerized diagnostic method that apply Inception V3 architecture. This method achieves an impressive accuracy, when tested on the HAM10000 dataset, highlighting its effectiveness in identifying and diagnosing skin ailments. The hybrid machine learning approach, when applied to a dataset of 3672 categorized pictures, produces a diagnosis accuracy of 99.80% on testing while achieved validation accuracy of 84.27%. This shows promise for improving the categorization of skin cancer and potentially leading to advancements in diagnosis, treatment, and mortality rates.

Challenges in Requirements Engineering for IoT Solutions

Muhammad Asgher Nadeem — 2024-03-01

The Internet of Things (IoT) defines a network of devices connected to the internet and sharing a massive amount of data between each other and a central location. Requirement Engineering is a widespread field of knowledge that is influenced by a large number of factors, from client’s participation to analyst’s expertise. In this paper, all the major problems in the field of requirement engineering are categorized on the basis of their area of origination. This paper categorizes these problems in three parts: the analyst’s, the client’s, and the generic problems. All the major problems of requirement engineering are separated in the above-mentioned categories. We are presenting the possible solutions for IOT issues and challenges.

A Hybrid Machine Learning Model to Predict Sentiment Analysis on X

Fiza Malik — 2024-03-01

Social media, particularly Twitter now ????, have emerged as pivotal arenas for sentiment analysis due to their pervasive nature and significant impact on shaping opinions. Our research delves into Roman-Urdu sentiment analysis within the burgeoning realm of social media, addressing a significant gap in research. Leveraging machine learning techniques, it emphasizes the scarcity of sentiment analysis studies in this linguistic domain, specifically on platforms like Twitter. The methodology involves meticulous data collection from English and Roman-Urdu tweets, followed by comprehensive preprocessing to refine and enhance dataset quality in python. Feature extraction retrieves key characteristics like subjectivity and polarity, enabling a nuanced sentiment analysis. Our technique evaluates precision, don't forget, F1 rating, and accuracy metrics the use of a complete evaluation framework on 4 machine learning classifiers: Naïve Bayes (NB), Random Forest (RF), Decision Tree (DT), and Support Vector Machine (SVM) algorithms. Roman Urdu sentiment analysis has advanced way to the results, which show how nicely this method works to classify all three sentiments (Hate, Offensive, and Neither) in multilingual social media content.

A Layered Analysis on Navigating the Landscape of IoT Attacks: A Survey

Rania Naveed — 2024-03-01

The Internet of Things (IoT) has become a focal point in contemporary research, capturing the imagination of researchers as they explore its transformative potential for the future. Despite the notable progress and development in the field of IoT, a host of vulnerabilities has surfaced, challenging the overall security of this technology. Interestingly, various attacks on IoT have been conceived even before its widespread commercial adoption. This study takes a deep dive into the realm of IoT attacks, categorizing them based on the layers of the IoT architecture. It aims to provide a comprehensive understanding of these attacks without delving into specific countermeasures. The given research study presents a state-of-the-art survey, shedding light on the diverse spectrum of attacks within the IoT framework, and contributing valuable insights to the ongoing discourse on IoT security.

AI-Powered Radiology: Enhancing Efficiency and Accuracy in Knee Osteoarthritis Diagnosis through Automated Bone Segmentation

Ayesha Kiran — 2024-03-01

A significant number of people experience a decline in their quality of life annually due to Knee Osteoarthritis, a debilitating joint condition. Clinicians commonly diagnose osteoarthritis by identifying potential joint space narrowing visible in knee X-ray images. As bone segmentation is crucial for accurate measurement of joint space width, this process require an automated solution in the form a U-Net model. This paper demonstrates a deep learning-driven method for automated joint detection and bone segmentation in knee radiographs, incorporating a U-Net model with VGG11 encoder. The proposed solution effectively detects and extracts joints from radiographic images. Additionally, it precisely segments bones, obtaining a segmentation mean Intersection over Union (IOU) score of 0.963. An algorithmic approach is introduced for measuring vertical distances to determine the joint space width between the femur and tibia bones. With an accuracy rate of 89%, the images are reliably classified as either normal or exhibiting osteoarthritis.

Diagnosis of Hepatitis Disease Classification Using Non-Linear Compound Algorithms

Aqsa Jameel — 2024-03-01

The liver is a vibrant organ of our body that cast-off to process nutrients, scrimmage infections, and blood baptism. Inflammation of the liver causes an acute or chronic infection called hepatitis which happens when body tissues are contaminated due to the devouring of ethanol, some medication, and toxin. Detecting disease in its early stages can be difficult, as practitioners often struggle to predict the disease due to its ambiguous symptoms. The primary objective of this study is to develop a model for predicting liver disease in its early stages, which will aid practitioners in accurately diagnosing hepatitis. Different algorithms and techniques are available for data mining to solve data discovery problems and arrangement. The discussed algorithms are supervised learning whose labels are defined, in which classification is the vital method. However, this study focuses on the comparison of Classification and Regression tree (CART) and Java 48 (J48) using a 10-fold cross-validation method with comprehensive medical accuracy and well-informed decisions for disease detection. Amid these algorithms, decision trees are the simplest and easiest algorithms for understanding, decision making due to hierarchal structure in nature. The data set used in this analysis consisted of 155 patients with two classes and performance measures among said models. The comparison and investigation of the results revealed that the J48 algorithm shows improved performance with the highest classification rate and performance measures over CART obtained as an accuracy of 80%, a sensitivity of 88%, and a specificity of 52% to quantify how good and reliable the test is at detecting a positive disease. This article will aid physicians in classifying high-risk patients by making a novel prognosis, fending off and managing the disease by allowing data analysis of different patients by minimizing the need for excessive testing. Consequently, it will improve the patient’s confidentiality by keeping them secure from health ramifications.

Boosting Early Diabetes Detection: An Ensemble Learning Approach with XGBoost and LightGBM

Faheem Mazhar — 2024-03-01

Given the increased prevalence of diabetes, early identification and prognosis of the condition are essential to avoiding long-term health consequences. Diabetes is an enduring medical illness that may have a role in the global health crises. The International Diabetes Federation estimates that 382 million people worldwide have diabetes. This number is expected to double by 2035, to reach 592 million. A medical condition known as diabetes is brought on by an excessively high blood glucose level. Diabetes is the main cause of renal failure, blindness, amputations, heart failure, and stroke. In order to develop a computerised approach for diabetes prediction, this work uses machine learning (ML) techniques on the Pima Indians dataset and private diabetes information. The aim of this project is to combine the findings from multiple machine learning techniques to create a system that can more accurately predict a patient's risk of developing diabetes in their early years. Techniques including logistic regression, SVM, RF, KNN, and decision trees are used. For every algorithm, the model's accuracy is computed. The model that predicts diabetes with the best accuracy is then chosen. We have achieved remarkable results in terms of accuracy, precision, recall, and F1-score for the models on the dataset by utilising several machine learning classifiers and putting feature removal techniques like feature permutation and hierarchical clustering into practice. This suggests that our characteristics or data are not limited to specific models.

Experimental Investigation of Peltier Based Thermoelectric Cooling System for Vaccine Storage

Yousaf Khan — 2024-03-01

Vaccines are typically administered during routine immunization programs, with a critical focus on maintaining a cold chain system. The cold chain system plays a pivotal role in ensuring the efficacy of vaccines. In this paper, a portable solar-powered vaccine carrier box based on the Peltier Effect for effective vaccine cooling is designed and experimented. The system is equipped with four 12V DC 3.5A Thermoelectric Cooler (TEC) Peltier Modules, with strategically positioned heat-sinks outside the cooling box. Additionally, the setup incorporates a 12V 180W Solar Panel for daytime power generation and eight Rechargeable Li-Po 3.7V 4500mAh batteries for uninterrupted operation for 3 hours. The experimentation involved three distinct conditions. Firstly, in the empty cooling box, a gradual temperature decrease was observed. Secondly, with the introduction of 10 sterile glass vaccine tubes filled with water, the temperature decreased more slowly, reaching 15°C in 62 minutes and 8°C after 90 minutes. Lastly, with 6 vaccine tubes, it took approximately 55 minutes to reach 15°C and about 90 minutes to achieve the desired 8°C temperature. The system exhibits a Coefficient of Performance COP of 0.42. The findings emphasize the effective cooling performance of the novel storage system, highlighting its ability to maintain temperatures below 15°C, a critical factor in preserving vaccines.

Diagnose of COVID-19 by Using CNN Based Models on Medical Images

Tauqeer Ahmed Khan — 2024-03-01

COVID-19 is a fast-spreading viral disease that affects both animals and people. Chest computed tomography and chest radiography are superior imaging techniques for detecting lung problems. This work was done to diagnose the COVID-19 disease by using CNN-based models on chest gray scale CT scan images. In the current study, the poster anterior view of the chest CT scan images has been used for both healthy subjects and patients with COVID-19 illness. Using the deep learning on CNN based models; we compared the performance of cleaned as well as augmented models. We have contrasted and compared precision of the Inception V3, Xception, and ResNeXt models. The dataset was generated from the Kaggle repository; there were 15102 gray scale chest CT scan pictures in the collected data set, including normal, COVID, and pneumonia. Total data set is further divided into training and validation sets. The Xception model detects images of chest CT scan with an accuracy of 98.90%, which is higher than state of the art approaches. This study makes medical claims and examines different classification schemes for patients infected with COVID-19.

A Technique for Safeguarding Legitimate Users from Media Access Control (MAC) Spoofing Attacks

Makhdoom Muhammad Naeem — 2024-03-01

Both wired and wireless networks are susceptible to MAC spoofing attacks, in which an attacker pretends to be a legitimate user on the network by changing the MAC addresses of both Ethernet and wireless devices. Authorized MAC addresses can, however, be easily impersonated to launch a variety of attacks, including phishing, denial-of-service attacks, SIP Registration hijacking attacks, and denial-of-access attacks. Cybercrime detectives face a formidable barrier when attempting to impersonate a legitimate user's MAC address. We suggest using Kea DHCPv4 and MySQL as the back-end database to implement an SMS-based system. Our objective is to completely safeguard real users from MAC spoofing assaults. The combination of MAC address and mobile number as host-name is what distinguishes the suggested technique from others. If the legitimate user is utilizing the network service, he can ignore the alarm message from the SMS application. If not, the legitimate user can tell the network administrator to block the IP address of the errant user. In this article, we successfully apply the suggested technique, identify the intruder via SMS alert message, and safeguard the machine of the authorized user via the network administrator. A prototype is used to confirm the veracity of the suggested methodology. The discussion section goes into further detail about how the proposed methodology was implemented using this prototype.

Sentiment Analysis of Diabetes Patients’ Experiences Using Machine Learning Techniques

Muhammad Shumail Naveed — 2024-03-01

Diabetes is a long-term medical disorder that affects blood sugar levels and can cause a variety of related issues, such as heart disease, kidney damage, nerve damage, eye damage and skin ailments. The impact of diabetes on patients' emotional sentiment has not been thoroughly studied, creating a gap in current knowledge on the potential psychological consequences of the disease. This study explores the connection between emotional sentiment and diabetes in an effort to close this knowledge gap. During this study, 215 online forum posts, including patient experiences, problems, routines, and suggestions, were analyzed using two widely used sentiment analysis models, TextBlob and Vader, to investigate whether diabetes affects patients' emotional state. The overall results indicate that diabetes may affect the sentiments of diabetic patients, as observed in their experiences, problems, and suggestions shared as posts on online discussion forums. However, it cannot be conclusively concluded that diabetes always has a significant and directly adverse impact on the sentiments and emotions of diabetic patients. To ascertain whether there is a link between emotions and diabetes, additional in-depth study on the sentiment analysis of patient experiences with diabetes is required, and to identify the specific circumstances under which this association may exist.

Mapping and Temporal Analysis of Wheat Crop Using Remote Sensing Imageries Burewala, Pakistan

Sidra Saeed Rana — 2024-03-01

Research is proposed to study the monitoring and detection of the wheat crop using remote sensing Imagery of Burewala Pakistan.The objectives of the study are to map agriculture land using Landsat-9 (USGS) Imageries and On-field visit of the crops. The image processing techniques such as Normalized Difference Vegetation Index (NDVI), Masking, were used to map the land cover changes. Further, the changes in land cover are correlated with reference to agricultural land.The proposed methodology and scientific approach are simple and effective and could be utilized for better management for the development of crops in the region. The aim of this research is related to wheat crops in Burewala using state of the art , remote sensing imageries and field data. We have deferment the socioeconomic impacts of wheat crop declining . The results would help to conclude the socioeconomic aspects related to the wheat crop productivity.

AI-Powered Customized Learning Paths: Transforming Data Administration For Students On Digital Platforms

Majid Ali — 2024-03-01

More than ever, effective information management and customized learning opportunities are needed for college students. This is because online education is being preferred by some people and thus choosing it. This article concentrates on AI-based personalized learning paths that the internet platforms are evolving of lately. Educational institutions can leverage AI algorithms and data analytics to find the unique learning characteristics of the student and design his/her educational route in a manner that will be most suitable for his/her learning style, academic achievements and memorization capabilities. AI-driven personalized learning concept is discussed in detail and the outline includes assignments, quizzes and peer assessed tests. Besides, paper argues on potential problems and ethical problems related to bias and data privacy. The main concern of this study is that personalized study paths lead to both the improvement of students' results in academic subjects, and also to the development of important skills of them for a digital era, by investigating this transformative power of AI in detail.

A Multiclassification Framework for Skin Cancer detection by the concatenation of Xception and ResNet101

Ahmad Naeem — 2024-03-01

Skin cancer is a deadly type of cancer that is responsible for millions of fatalities annually across the globe. This malignant cancer occurred due to the proliferation of abnormal epidermal cells, which subsequently spread to adjacent tissues and spread to other organs and tissues through the lymph nodes. Changes in lifestyle and sun-seeking behaviors have contributed to the increase in the incidence of skin cancer. It is critical to accurately identify and classify skin cancer to prevent the serious effects that result from delaying detection and treatment. This research paper presents a newly developed deep learning model that makes use of two advanced artificial intelligence techniques, Xception and ResNet101. This method attains an extraordinarily high degree of accuracy by using the special advantages of two strong networks. The Xception-ResNet101 (X_R101) model is capable of differentiating specific categories of skin cancers, such as melanoma (Mel), melanocytic nevus (Mn), basal cell carcinoma (bcc), squamous cell carcinoma (Scc), benign keratosis (Bk), Actinic keratosis (Ak), Dermatofibroma (Df) and Vascular lesion (Vl). The implementation of borderline SMOTE improves performance substantially. A comparison is performed between the proposed methodology and four benchmark classifiers: MobileNetV2 (BM3), DenseNet201 (BM4), InceptionV3 (BM1), and ResNet50 (BM2) and state-of-the-art classifiers. To evaluate performance of the proposed methodology, three publicly available datasets (PH2, DermPK and HAM10000) are utilized. The X_R101 model attains a prediction accuracy of 98.21%. The method's accuracy and effectiveness provide benefits to dermatologists and other healthcare practitioners in terms of timely identification of skin cancer.

Diagnosis the Bearings Faults through Exercising Deep Learning Algorithms

Iftikhar Ahmad — 2024-03-01

Rolling bearings are vital components in process industries, and their faults can disrupt industrial processes. This paper presents a novel approach to diagnose the health state of rolling bearings, combining time-frequency domain signal analysis with deep learning models, namely ResNet-50 and DenseNet-121. Utilizing a dataset from the CWRU bearing datacenter containing various bearing health conditions, including normal and faulty states, this research addresses the limitations of conventional statistical feature-based fault detection methods. These scalograms are converted into grayscale images to optimize the learning process. The final grayscale CWT images are fed into the deep learning models for fault classification. Results indicate that the proposed framework, particularly the combination of CWTSV and DenseNet-121, yields promising outcomes of 95.25% and 99.77% respectively, surpassing existing methods for rolling bearing fault diagnosis. This approach holds potential for significantly enhancing industrial maintenance practices and ensuring process reliability.

Air-Sense: Enhancing Crop Yield and Quality through Integrated IoT-based Air Quality Monitoring System

Faseeha Munir — 2024-03-01

Air quality has a significant effect on crop production and quality. Crops are affected when they are exposed to certain air pollutants. The effect might range from visible spots on the leaves to reduced growth and yield, and plant’s early maturity. The degree of the damage was determined by the concentration of the specific pollutants and several other variables. This paper presents Air-Sense that is an IoT-base real-time air quality monitoring system. It is not only monitoring the quality of air and weather parameters around the crop but also deter-mine its effect on the crop leaves. Air-Sense monitors four gases: hydrogen gas (H₂), Carbon Monoxide (CO), Ozone (O₃) and Carbon dioxide (CO₂). The dust particles (PM2.5), s air temperature, soil moisture and humidity are also determined with this system. The system has been evaluated on the cotton crop by measuring leaf chlorophyll concentration with the help of the image processing model with an average validation accuracy of 93.47% based on an image dataset of 2000 healthy and infected cotton leaves. This system will enable farmers to monitor their fields air quality and make timely decisions about the best suitable sowing location for crops to get better yield and quality.

Convolutional Approaches in Transfer Learning for Facial Emotion Analysis

Ahmed Faraz — 2024-03-01

The scientific community has shown significant interest in facial emotion recognition (FER) due to its possible applications. The primary function of Facial Expression Recognition (FER) is to associate various facial expressions with their respective emotional states. Feature extraction and emotion recognition are the primary constituents of conventional FER. The inherent feature extraction capabilities of Deep Neural Networks, particularly Convolutional Neural Networks (CNNs), have resulted in their extensive utilization in Facial Expression Recognition (FER) currently. While previous studies have explored the utilization of multi-layer shallow convolutional neural networks (CNNs) for addressing facial expression recognition (FER) tasks, a significant drawback of these models is their limited capacity to accurately extract features from high-resolution photos. Many of the existing methods also exclude profile views, which are crucial for real-world facial expression recognition (FER) systems, in favor of frontal photographs. This research introduces a highly complex Convolutional Neural Network (CNN) model that incorporates Transfer Learning (TL) to enhance the precision of Facial Expression Recognition (FER). The proposed approach for satisfying the FER criteria involves utilizing a pre-trained DCNN model and fine-tuning it with facial expression data. Subsequently, it substitutes the dense higher layer(s) of the model. To improve the precision of Facial Expression Recognition (FER), a new approach is claimed that consists of iteratively applying the fine-tuning technique on each of the pre-trained DCNN blocks. The validation Facial Expression Recognition FER system of eight DCNN models trained previously, especially VGG-16 and VGG-19 is undertaken. The authentification process is run through two data sets, namely KDEF and JAFFE. But as tricky as the FER is, including the various perspective analysis which is part of the KDEF dataset, the proposed approach is a stand out in the high accuracy that it records. VGG-16 achieved the highest FER accuracies of 93.7% on the KDEF test set and 100% on the JAFFE test set using a 10-fold cross-validation. The assessment emphasizes the benefits of the proposed Facial Emotion Recognition (FER) system, particularly in its ability to reliably detect emotions. It demonstrates promising outcomes on the KDEF dataset, specifically in the context of profile views.

Predictive Modeling of Chronic Kidney Disease Using Extra Tree Classifier: A Comparative Analysis with Traditional Methods

Muhammad Imran — 2024-03-01

With a high prevalence of morbidity and death, chronic renal illness is a major global health concern. Conventional diagnostic methods frequently miss the disease until it has grown to an advanced stage, despite the fact that prompt diagnosis and treatment can greatly improve patient outcomes. This work suggests a unique method for utilizing machine learning (ML) algorithms to identify kidney sickness, which might offer a solution to this urgent healthcare problem. One of the many industries where machine learning—a subset of artificial intelligence—has demonstrated great potential is healthcare. Due of its capacity to make predictions and Take note of the data, it is a useful tool for predicting illnesses. This study uses a variety of clinical indicators along with machine learning methods to predict when chronic kidney disease (CKD) will manifest. The proposed model uses a dataset comprising numerous patient records with various attributes. These attributes are used as input features for the machine learning algorithm. The target variable is the presence or absence of chronic kidney disease. A number of machine learning algorithms are used, and their performances are contrasted, including KNN, Chronic Kidney Disease, Machine Learning, Gradient Boosting Classifier, Ada Boost Classifier, Random Forest Classifier, XgBoost, Cat Boost, and Extra Trees Classifier. To assess each algorithm's predictive accuracy, sensitivity, specificity, and other performance metrics, a subset of the dataset is used for training, and afterwards the algorithm is tested using untested data. The findings show that machine learning algorithms, some of which are more accurate than others, can predict chronic kidney disease. According to these results, machine learning may prove to be a useful tool in the early diagnosis of chronic kidney disease, allowing for prompt intervention and maybe leading to better patient outcomes.

Enhancing Rumor Detection on Social Media Using Machine Learning and Empath Features

Hafiza Anum — 2024-03-01

In today's society, social media serves as a significant platform for information sharing, especially during news events when real-time updates are provided. Its accessibility, speed, and simplicity make it a valuable source of firsthand knowledge, enabling individuals to stay informed and connected, even during disasters. However, alongside its benefits, social media also harbors misinformation or rumors, which spread rapidly and can have detrimental effects. These rumors, unverified statements circulating on social platforms, can hinder the effectiveness of social media, particularly during crises, by disseminating false information and impeding real-time assistance efforts. Various approaches, including manual and automated classification models, have been employed to identify and address rumors on social media. While many existing methods focus on known rumor stories and predefined features, our research adopts a novel top-down approach that considers real-time tweets. We propose training multiple machine learning algorithms using an empath method to automatically extract additional features for classifying rumors and non-rumors. By incorporating these features, we aim to enhance the accuracy of rumor detection compared to previous methodologies, ultimately improving the efficacy of social media in disseminating reliable information.

Sales Forecasting Using Machine Learning Algorithm in the Retail Sector

Saira Malik — 2024-03-01

The ability to predict future sales is essential for modern firms. The already difficult work of sales forecasting is made much more difficult by a lack of data, missing data values, or outliers. Regression has a stronger relationship with sales forecasting complexity than time series does. Intricate patterns in the dynamics of sales that also involve a range of risk factors can be discovered using machine learning algorithms and supervised machine learning techniques. A company's sales projections need to be correct for it to succeed. By utilizing a reliable sales projection model, businesses may identify potential risks and make smarter decisions. In this study, Rossmann sales data will be analyzed using the Extreme Gradient Boosting (XG-Boost), FB-Prophet, and autoregressive integrated moving average (ARIMA) prediction models. A corporation can reduce costs associated with excess inventory, make future plans, and boost profitability with the use of an accurate sales forecast. Therefore, the model needs to be assessed using statistical techniques like R2, RMSE, and MAE. To determine if models are more accurate at predicting sales, the results are employed.

Securing Cloud Environments: A Convolutional Neural Network (CNN) approach to Intrusion Detection System

Syed Younus Ali — 2024-03-01

Cloud-computing has become an essential portion of recent IT structure, contribution scalable resources and on-demand services to users. However, the increasing reliance on cloud environments has raised concerns about security, especially with the rise of sophisticated cyber threats. Intrusion detection systems (IDS) play a crucial role in detecting and mitigating possible security breaches. In this studies proposes a approach to enhance intrusion detection in cloud computing through the CNN. This deep learning architecture adapted for the unique challenges in cloud computing security. Unlike traditional IDS methods that rely on rule-based or signature-based approaches, the CNN-based intrusion detection system presented in this research leverages the network's capability to automatically learn hierarchical features from raw data. This study is involves the collection of diverse and representative datasets from cloud environments, including normal network traffic and various types of attacks. The CNN is trained on these datasets to learn the inherent patterns of legitimate activities and deviations indicative of potential intrusions. The proposed system demonstrates its adaptability to evolving threats by continuously updating its knowledge through regular retraining with new data. The evaluation of the CNN-based intrusion detection system is conducted through comprehensive experiments, comparing its performance against traditional methods. The results indicate that the CNN-based approach outperforms conventional IDS techniques, demonstrating its potential as a robust and efficient solution for intrusion detection in cloud computing environments.

Pitch Control of Variable Speed Wind Turbine Through PI Controller

Naila Ramzan — 2024-03-01

Recently, due to energy shortages and environmental concerns, considerable attention has been paid to renewable energy, especially wind energy. As the penetration of wind energy into the electrical power grid is increased significantly, the effect of wind turbine systems on frequency and voltage stability is becoming ever more important. Pitch angle control is typically used to control the wind energy system's output torque if wind speed reaches base speed, and other variables such as wind speed, generator speed, and generator power can also be chosen for control. In this paper a pitch control scheme is proposed which is based on a PI controller design. Comparative analysis of the proposed strategy with that of already existing ones shows that it gives an overall better result than those prevalent techniques.

Citation Count Prediction of Scholarly Articles

Lubna Zafar — 2024-03-01

Assessing the citation count is crucial for gauging the impact of scientific publications. Predicting future citation counts can assist researchers in discovering references and delineating research areas. In our study, we introduce a novel model called FoS Trend based Citation Count Prediction (FTCCP), which aims to forecast the citation count of scientific articles by leveraging field of study (FoS) trends and early citation counts. By analyzing the citation patterns within the first few years post-publication (specifically 1-3 years and 1-5 years), FTCCP extrapolates the long-term citation impact of an article. Notably, we focus solely on the FoS trend and Early Citation Count without considering other factors such as authorship, publication venue, or journal. While some prior research incorporates a broader range of features for citation prediction, we intentionally keep our model simple to ensure its applicability across diverse research domains.
Our investigation revolves around two feature categories for FTCCP: FoS trend and Early Citation Count. We employ Multiple Linear Regression to develop the citation count prediction model. Results from experiments conducted on the Microsoft Academic Graph (MAG) dataset demonstrate promising outcomes, indicating the effectiveness of FTCCP when utilizing FoS trend and Early Citation Count compared to models relying solely on citation history, as evidenced by higher R² scores. Furthermore, our proposed features exhibit superior performance compared to traditional ones.

Forensic Analysis of Modern E-Business Applications

Muhammad Asim — 2024-03-01

In this paper, a plugin is developed a for our automated digital forensics framework to extract and preserve the evidence from the IOS-based mobile phone application, Olx. This plugin extracts personal details from Olx users, e.g, user name, mobile number, User Location, Country name, State name, City name, Last check-in attempt, Ad Images between different Olx users. While developing the plugin, we identified resources available in IOS-based devices holding key forensics artifacts. We highlighted the poor privacy scheme employed by Olx. This work has shown how the sensitive data posted in the Olx mobile application can easily be reconstructed, and how the traces, as well as the URL links of visual messages, can be used to access the privacy of any Olx user without any critical credential verification. We also employed the anti-forensics method on the Olx IOS application and were able to restore the application from the altered or corrupted database file, which any criminal mind can use to set up or trap someone else. The outcome of this research is a plugin for our digital forensics ready framework software which could be used by law enforcement and regulatory agencies to reconstruct the digital evidence available in the Olx mobile application directories on IOS-based mobile phones.

Social Media Platform Prediction for Digital Marketing using Machine Learning Techniques

Fazli Mola Jan — 2024-03-01

Every business wants to connect with their users in seamless and efficient manner. Social media play vital role in achieving that goals. Social media plays vital role in the marketing of any business. But on the other side, there are so many social popular media platforms and it is difficult for businesses to choose which social media platform will perform better for them. This study focuses on choosing best social media platform for digital marketing using machine learning techniques. We studies various algorithms and then after careful consideration we pick Random Forest algorithm for choosing optimal social media platform. Machine learning is playing very important role in making informed decisions and we are taking advantage of that in the shape of utilizing past data related to various business categories to generate most accurate results. To train the model and make accurate results, we collected data from 10,000 businesses having various parameters like business category, location, audience demographics, and past advertisement data. After making dataset, we preprocessed and cleaned the data and to obtain better results from the model. We trained our model on 70% of the data and then tested it by 30% of the remaining data. Our model shows 77% accuracy in choosing social media platform for promoting their businesses. We also made a mobile application for the businesses so that they can use it and predict the best social media platform for promoting their business.

The Sentence Level Sentiment Analysis of Cyber Trolling Tweets Using Machine Learning Technique

Wareesa Sharif — 2024-03-01

People are utilizing the networking site Twitter not only for social interaction but also to express their opinions, thoughts, news, and personal information in the form of text, videos, and pictures. Many of these tweets are cyber-trolling-related, psychologically devastating, and should be on the notice of the police. However, analyzing these tweets manually is quite difficult. Therefore, an intelligent mechanism is required to examine and polarize those cyber trolling-related tweets. Thus, in this paper, Valence Aware Dictionary Sentence (VADsentence) Miner has been proposed to perform Sentence Level Sentiment Analysis (SLSA) using machine learning (ML) techniques. For this purpose, tweets are pre-processed and sentences are extracted on the base of adjectives, adverbs and noun phrases. For SLSA, a combination of lexicon and rule-based approach named Valence Aware Dictionary and Sentiment Reasoner (VADER) is used to compute the sentiment polarity of tweets based on sentences. The proposed VADsentence Mines experimented with the feature selection technique TF-IDF and machine learning algorithms. Results of VADsentence Miner are compared with TextBlob in that VADsentence Miner outperformed 90% in accuracy, 82% in precision, 74% in recall, and 78% in F1-score on the Random Forest machine learning classifier and Term Frequency Inverse Document Frequency (TF-IDF). Textblob however, could archive 67% of accuracy on Random Forest and Term Frequency Inverse document frequency (TF-IDF).

Breast Cancer Diagnosis by Exploiting the Permutations of Principal Components by Ensemble Classification

Aimen Sikander — 2024-03-01

In many breast cancer computer-aided diagnosis problems with larger feature dimensions and fewer feature instances, the classification does not get optimal training. This is because a decision boundary is represented by the number of parameters directly proportional to the feature dimensions. Since the optimal training of such high-dimensional features requires a large training set. Unluckily, if the training set is not sufficiently large to generate good n/l ratio, the training results in an ineffective and inefficient classification model. To resolve the problem of large dimensions, the conventional employment of feature reduction techniques results in efficient training however it yields the degraded classification performance. In this paper, we consider this problem to have effective and efficient training in large dimensional datasets when the available dataset is not sufficiently large. For this purpose, we hybridize principal component analysis with ensemble classification. For this, different combinations of principal dimensions have been determined by the concept of power sets in mathematics. A dedicated base learner then exploits each principal dimension combination. Then, all these base learners are combined to construct a hybrid ensemble principal component analysis-based classifier, Ens-PCA. The proposed Ens-PCA technique is tested using Wisconsin diagnostic breast cancer (WDBC) data set and the results show its outperformance as compared to the contemporary principal component analysis and ensemble classification techniques.

Analytical Study of OLTP Workload Management in Database Management System

Shahid Rafique — 2024-03-01

This study determines the importance of enlightening OLTP library management systems to provide a fast and unified experience to users. Using procedures such as query optimization, caching, database indexing, and code-level developments, the study pointedly recovers transaction processing speed and overall system performance. Scalability testing and user feedback sustenance these results. Performance should be monitored frequently and modifications made to confirm everything continues to run as predictable. Moreover, additional enhancements may be possible in the future. This experiment taught us how to improve the functionality of today's critical OLTP library systems. By using smart technology, we have gone beyond our original goal of simplifying the lives of library staff and patrons. The booming success of this experiment prompts us how significant flexibility and continuous development of technology are. In this way, our library system can assist as a cooperative resource that varies with the requirements of the public. As the study moves forward, the perceptions expanded from this experiment will not only help us reinforce our systems but also help us apply new concepts to improve customer satisfaction.

Computer Vision Based Solanum Lycopersicon Leaf Disease Detection Using Transfer Learning

Abdul Hakeem — 2024-03-01

Solanum Lycopersicon, mostly known as tomatoes, are one of the most essential and extensively consumed crops, with yields varying depending on cultivation methods. Tomato leaf disease is the most critical factor in both supply and quality of tomato crops. As a result, it is essential to identify and determine these diseases appropriately. Different diseases can impact the production of tomatoes, and early detection is crucial in reducing the consequences and encouraging a healthy crop yield. Improved approaches for disease detection and classification have been widely used. Various studies have been proposed to identify tomato leaf diseases, but they must be enhanced due to their accuracy and effectiveness as trained on the limited dataset. This work aims to support farmers in accurately diagnosing early-stage tomato leaf diseases namely: Bacterial spot, leaf spot, blight, curl virus, leaf mold diseases and delivering necessary information. In this study, deep learning-based models applied using a transfer learning technique. Different models, such as ResNet-50, VGG-16, and VGG- 19 are applied. Python is used to train deep learning models using Google Colab. The dataset acquired from publicly available repositories, namely, Kaggle. The image data is preprocessed using resizing and rescaling. The applied models will be evaluated in accuracy and performance to choose the best one.

Deep Learning for COVID-19 Diagnosis Using Pretrained and Non-Pretrained Models

Muhammad Basit Umair — 2024-03-01

This article proposes a deep-learning approach to classify COVID-19 cases using image data. Our model uses a convolutional neural network (CNN) to extract features from chest X-rays and classify them as positive or negative for COVID-19. A COVID-19 case dataset is compared to traditional machine learning methods to evaluate model performance. The results obtained demonstrate the effectiveness of the deep learning model in accurately detecting COVID-19 cases with an overall accuracy of 96%. This approach is helpful for rapid and automated diagnosis of COVID-19, especially in resource-limited settings. The proposed method yielded remarkable results compared with recent results.

Analyzing Paper Citation Trend of Popular Research Fields

Lubna Zafar — 2024-03-01

The ever-expanding volume and diversity of scientific literature pose a significant challenge for researchers in detecting emerging, current, and future research trends. A trend represents the prevailing direction of research within a defined timeframe. Detecting trends involves identifying areas of growing interest over time, while trend analysis involves gathering data and discerning patterns. Despite the utilization of diverse methods for analyzing and identifying trends in scientific research, there remains a lack of comprehensive understanding regarding the significance of following research trends for citation of research papers. The objective of this research is to examine the significance of monitoring trends in Computer Science (CS) research, the influence of aligning with these trends on paper citations, and the correlation in citation patterns among papers within the CS domain. We analyze trends in CS conference papers and the evolution of research fields from 1985 to 2017 using the Microsoft Academic Graph (MAG) dataset of CS papers in the L1 field of study (FoS). Our experimental findings reveal that Data Mining, Artificial Intelligence, Computer Vision, Machine Learning, and Database research exhibit the highest publication trends. Additionally, our results suggest that papers within the same field demonstrate similar citation trends.

Enhancing LAN Security by Mitigating Credential Threats via HTTP Packet Analysis with Wireshark

Altaf Hussain — 2024-03-01

The world is connected digitally and the security of Local Area Networks has been dangerous increasingly. A process that is especially designed to secure networks from different ouside attacks is known as Cybersecurity. In this article, a Local Area Network threat scenario is discovered especially concentrating on the extraction of credentials by capturing the Hypertext Transfer Protocol packets. Most of the time, Local Area Network can be said a secure, but sometimes itcan have many vulnerablities to cybersecurity threats. An attacker can be connected with Local Area Network and using packets capturing and network anlysing tool wireshark; they can exploit the Hypertext Transfer Protocol vulnerabilities to obtain login credentials. Attackers have motives to get specific credentials, they perform actions to get IP addresses, emal addresses being used in communication and financial details, by network trffic examination. Differenet protocols such as Hypertext Transfer Protocol, Address Resolution Protocol, and Transmission Control Protocol are can be captured and analyzed by this tool. To get filtered packets, Wireshark provides the best filtering selections and interpret into packets. For the security of Local Area Network, implementation of various security approaches including encryption of data and protocols, Firewall, IDS/IPS implementation, network segmentation, ethernet cables usage, use of Hypertext Transfer Protocol Secure and Multifactor authentication is deployed. Network traffic should be monitored, apply port security, and allow only registered Media Access Control in Access Point. The proposed solution enhanced the security of the Local Area Network and mitigated the cybersecurity threats. Network and connected devices monitoring regularly and activity of traffic packet-capturing tools can make the Local Area Network more secure.