Lahore Garrison University Research Journal of Computer Science and Information Technology
Not a member yet
227 research outputs found
Sort by
Deep transfer CNNs models performance evaluation using unbalanced histopathological breast cancer dataset
Full text linkCancer is one of the top deadly diseases. Of this disease, around about 9.8 million death cause annually. It has been recorded by the American Cancer Society that every eight women die due to breast cancer in the USA. In this paper, we have identified eight different lesion categories: Benign Tumor: Adenosis-Adenoma, Fibro-Adenoma, Phyllodes-Tumor, Tubular-Adenoma, and Malignant Tumor; Ductal-Carcinoma, Lobular-Carcinoma, Mucinous-Carcinoma, Papillary-Carcinoma. The main contribution of this paper is to examine the performance of five pre-trained CNN models on an unbalanced cancer dataset for cancer prediction. The identification of different cancer tumors has been recognized by using transfer learning models namely ResNet50, ResNet101, ResNet152, VGG16, and VGG19. BreakHis dataset has four different magnifications (40x-100x-200x-400x), and used for experiments setup in this study. The total number of images for all magnification levels is 7909. The experimental results state that the pre-trained model Residual Net with different variations has worked well 91%~96% as compared to other pre-trained models. The ResNet101 architecture model has gained a multiclass identification higher than 95%. In this paper, the proposed methodology has different evaluation parameters such as accuracy, recall, and f1-score of all pre-trained models that will help to build optimal, and automated breast lesion multiclass identification
DIGITIZING TVET EDUCATION THROUGH UNIFIED ANALYSIS OF PERSONALITY AND LEARNING SKILLS
Full text linkPersonality exerts a pivotal influence on Technical and Vocational Education and Training (TVET) by shaping individuals' learning approaches, social interactions, and practical skill applications. Customizing educational strategies and curriculum designs to suit diverse personality profiles cultivates a more conducive learning environment, enabling each student to optimize their abilities and contribute effectively to their chosen vocational fields. Incorporating personality considerations into TVET programs not only enhances skill acquisition but also fosters personal growth and professional success among learners. This research aims to test our hypothesis that can we predict TVET courses based on personality traits and whether is there any role of demographics (age and gender), and examination performance scores in TVET course prediction. Data for this research were collected from one of the largest TVET training providers and a five-fold cross-validation technique with MDS analysis and Decision Tree methods were used. The result discovered that TVET courses can be predicted based on personality traits and demographics and Examination Scores have a significant role in TVET course prediction. Prediction accuracy of 78% is achieved by SVC, 76% by Naïve Bayes, and 74% by Random Forest classifies
Use of Code Refactoring Transformation in Software Advancement
Full text linkIn this paper, discuss the refactoring object-oriented code effects on software quality. We can say that refactoring is the reengineering process of code in software development, sometimes it is very helpful because it is not as expensive. The external behavior of software will not be changed, that is the big advantage of code refactoring. A lot of different researcher’s research on it to evaluate the results and check the software quality because companies do not compromise on quality standard. In this paper, refactoring technique shift to software requirements because if requirements are clear and definite then all the development will be good and up to the mark. We will check it at an earlier phase of software development so in this sense it is not very expensive. Different five research questions and their answers are also part of this paper. We will use requirements divided approach in it. Staring with requirements gathering then filter, break, prioritize, numbering and then convert it into requirement document. Apply this approach to a case study (Hotel Management System) and derive the results at each phase
Classification of Microscopic Malaria Parasitized Images Using Deep Learning Feature Fusion
Full text linkAn infectious disease that causes a chronic and potentially life-threatening infection caused by microorganisms of the Plasmodium class, is malaria, or malarial disease. It is critical to detect the presence of Malaria parasites as early as possible to ensure that antimalarial treatment is adequate to cure the particular type of Plasmodium. This is to reduce death rates and to focus on various infections in the event of an adverse outcome. The purpose of this study was to develop an artificial intelligence approach capable of separating parasitized erythrocytes from normal basophilic erythrocytes as well as platelets overlying the red blood cells to overcome the high cost of Ma-laria diagnostic equipment. The tone and texture characteristics of erythrocyte images were extracted using histo-gram thresholds and watershed methods, and then fused with Squeeze Net and ShuffleNet algorithms. The measures included planning, preparing, approving, and testing Deep Convolution Neural Network Segmentation without preparation using a graphic processor unit. A total of 96 percent accuracy and specificity was obtained for the position of malaria in red blood cells based on the results of all of the tests. It has been demonstrated that deep learning can be effective in the field of clinical pathology. This provides new directions for development as well as increasing awareness of researchers in this field
A Fuzzy Clustering-based Approach for Classifying COVID-19 Patients by Age and Early Symptom Indicators
Full text linkThe devastating illness known as Covid-19 has disrupted the lives of individuals all over the globe and left a trail of devastation in its wake. The fact that we are unable to determine the severity of illness (SOI) class of the patient during the early stages of infection is without a doubt the most challenging aspect of this disease. An accurate classifier model has to be constructed in order to ensure that patients diagnosed with Covid-19 get prompt and individualized therapy. Within the scope of this investigation, we propose a useful fuzzy clustering based model for categorizing Covid-19 patients according to their age and the severity of their early symptoms (fever, dry cough, breathing difficulties, headache, smell, and taste disturbance). This method is superior to previous hard clustering tactics in terms of reducing the number of deaths that occur among patients suffering from coronavirus and increasing the likelihood that they will recover fully
Classifying Tweets with Keras and TensorFlow using RNN (Bi-LSTM)
Full text linkUnderstanding public opinion, sentiment analysis, and subject recognition have all become more and more important as social media platforms have grown exponentially. The methodology for categorizing tweets using Keras and TensorFlow with a Recurrent Neural Network (RNN) architecture, specifically Long Short-Term Memory (LSTM) units, is presented in this research article. The method uses word embeddings and other properties to improve tweet representation, allowing the model to reliably identify specified categories and capture contextual connections. Our RNN-LSTM model beats baseline methods after extensive testing and evaluation, proving its suitability for tweet classification applications. The model's comprehension of tweet content is further improved by the incorporation of pre-trained word embeddings as well as features like emotion scores and hashtags. The approach offers a thorough framework for using deep learning methods in tweet classification, opening the door for uses cases including sentiment analysis, topic recognition, and opinion mining. By providing knowledge on the possibilities of RNN-LSTM models and their use in comprehending and analysing social media data, this research makes a contribution to the area. The results emphasise how crucial it is to take temporal dynamics and contextual factors into account while handling tweet classification jobs. Future research may concentrate on researching other pre-trained embeddings, investigating advanced RNN architectures, and solving issues with noisy and biassed twitter data. Overall, the large volume of information published on social networking sites like Twitter may now be better understood and analysed thanks to this research.
 
EVALUATION OF PSNR VALUE FOR IMAGE SUPER-RESOLUTION USING DEEP LEARNING
Full text linkDeep learning is a branch of study that simulates the human brain and allows learning from large data. In recent years, deep learning techniques have become more prevalent in a range of applications. These techniques have been widely employed in image processing, especially for object detection, image augmentation, and super-resolution. Single-image super-resolution, which converts low-resolution images into high-resolution ones, is a significant application of deep learning. Outstanding perceptual and antagonistic outcomes have been achieved using multiple single-image super-resolution models. In this study, several noteworthy methods, including Super Resolution with Convolutional Neural Network, Super Resolution Residual Network, Super Resolution Generative Adversarial Network, and Enhanced Super Resolution Generative Adversarial Network are used for single image super-resolution. Each layer in these deep learning models carries out certain operations on low-resolution images to transform them into high-resolution equivalents. As a result, many high-resolution images can be produced from a single low-resolution image. We created a dataset of random images for our research. The original low-resolution images are compared to the high-resolution images that were created using a deep learning technique. Peak Signal-to-Noise Ratio and the Structural Similarity Index are used in the comparison. Additionally, the evaluation process includes computing the ground truth value. The highest similarity indexed is achieved by VGG54 with 96% of similarity.
 
Identification of Finger Vein Images with Deep Neural Networks
Full text linkTo establish identification, individuals often utilize biometrics so that their identity cannot be exploited without their consent. Collecting biometric data is getting easier. Existing smartphones and other intelligent technologies can discreetly acquire biometric information. Authentication through finger vein imaging is a biometric identification technique based on a vein pattern visible under finger's skin. Veins are safeguarded by the epidermis and cannot be duplicated. This research focuses on the consistent characteristics of veins in fingers. We collected invariant characteristics from several cutting-edge deep learning techniques before classifying them using multiclass SVM. We used publicly available image datasets of finger veins for this purpose. Several assessment criteria and a comparison of different deep learning approaches were used to characterize the performance and efficiency of these models on the SDUMLA-HMT dataset. 
Cloud Computing Services and Security Challenges: A Review
Full text linkAn architecture of computing that provides services over the internet on the demand and desires of users that pay for the accessible resources that are shared is refer as the cloud computing. These resources are shared over the cloud and users do not have to acquire them physically. Some of the shared resources are: software, hardware, networks, services, applications and servers. Almost every industry from hospitals to education is moving towards the cloud for storage of data because of managing the effective cost and time of organizing the resources physically on their space. Storage of data over the data centers provided in the form of clouds is the key service of the cloud computing. Users store their desired data on clouds that are publicly available over the internet and away from their boundaries in cost effective manner. Therefore, techniques like encryption is used for obscuring the user’s information before uploading or storing to the shared cloud devices. The main aim of the techniques is to provide security to the data of users from unauthorized and malicious intrusions
Diabetes Diagnosis through Machine Learning: An Analysis of Classification Algorithms
Full text linkDiabetes is a serious and chronic disease characterized by high levels of sugar in the blood. If left untreated, it can lead to numerous complications. In the past, diagnosing diabetes required a visit to a diagnostic center and consultation with a doctor. However, the use of machine learning can help to identify the disease earlier and more accurately. This study aimed to create a model that can accurately predict the likelihood of diabetes in patients using three machine learning classification algorithms: Logistic Regression (LR), Decision Tree (DT), and Naive Bayes (NB). The model was tested on the Pima Indians Diabetes Database (PIDD) from the UCI machine learning repository and the performance of the algorithms was evaluated using various metrics such as accuracy, precision, F-measure, and recall. The results showed that Logistic Regression had the highest accuracy at 71.39% outperforming the other algorithms