ICITR - 2023

Permanent URI for this collectionhttp://192.248.9.226/handle/123/22075

Browse

Now showing 1 - 20 of 47

item: Conference-Full-text
8th International Conference in Information Technology Research 2023 (Per Text)
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
item: Conference-Full-text
Using multispectral uav imagery for marine debris detection in Sri Lanka
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Velayuthan, P; Piyathilake, V; Athapaththu, K; Sandaruwan, D; Sayakkara, AP; Hettiarachchi, H; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
Marine pollution is a significant issue in Sri Lanka, with the country being a major contributor to marine debris. Marine pollution has the potential to adversely impact marine and coastal biodiversity, as well as the fishing and tourism industries. Current methods for monitoring marine debris involve labor-intensive approaches, such as visual surveys conducted from boats or aircraft, beach clean-ups, and underwater transects by divers. However, an emerging trend in many countries is the use of Unmanned Aerial Vehicle (UAV) imagery for monitoring marine debris due to its advantages, including reduced labour requirements, higher spatial resolution, and cost-effectiveness. The work presented in this study utilizes multispectral UAV imagery to monitor marine debris in a coastal area of Ambalangoda, Sri Lanka. For the automated detection of marine debris in captured images, this work replicates the state-of-the-art CutPaste method for region detection and utilized the ResNet-18 model with Faster R-CNN for the final classification of marine debris instances. The implemented approach demonstrated a classification accuracy of approximately 60% in automatic marine debris detection, laying the groundwork for potential enhancements in the future.
item: Conference-Full-text
Dominant color palette extraction in resumes using the new color pixel quantifier algorithm
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Perera, NN; Warusawithana, SP; Weerasinghe, RL; Hindakaraldeniya, TM; Ganegoda, GU; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
In the realm of resume analysis and enhancement, the extraction of dominant color palettes plays a pivotal role in assessing the visual impact of resumes. Existing methods designed for images with extensive color ranges have proven to be suboptimal when applied to the distinct context of resumes, which inherently possess a limited color palette. This paper introduces a novel approach that addresses this challenge effectively and efficiently. By minimizing the time required for palette extraction without compromising accuracy, the proposed method offers a practical solution for resume feedback systems. It is important to clarify that this research neither rejects nor supports existing methods; instead, it presents an alternative, tailor-made solution for resume analysis. In summary, this paper sets a promising precedent for more streamlined and functional dominant color palette extraction methods in the context of resumes, promising advancements in resume analysis and improvement.
item: Conference-Full-text
Predicting the performance of electrical machines using machine learning
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Manohar, VJ; Jha, SK; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
Electrical machines play an important role in our day-to-day life. Electric machines like DC motors and 3- phase induction motors are essential systems and widely used in domestic, industrial and transportation systems. In order to operate the machines optimally and efficiently, in real time operations, it is required to predict the performance parameters at various loaded conditions. With the advancements in the field of predictive modelling and analytics, several researchers have applied in the area of energy consumption prediction, fault prediction, weather prediction, power grid management and so on. In this paper, the machine learning techniques are demonstrated that may be used to examine the performance of electrical machinery by forecasting performance characteristics like speed and efficiency. To validate the performance of the predictive model, an experiment was conducted at the laboratory on dc motor and 3- phase induction motor to generate the required dataset to train the regression algorithms. The model evaluation metrics such MSE and the R2 value showed that the model efficiently predicted the performance of the electrical machines.
item: Conference-Full-text
Personal loan default prediction and impact analysis of debt-to-income ratio
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Rodrigo, KLS; Sandanayake, TC; Silva, ATP; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
Loan defaults affect the financial sector, particularly impacting banks and lending institutions, resulting in a rise of non-performing assets and financial strain. To counteract this trend, traditional credit assessments use methods like credit scores and exploitation of socio-demographic composition of the customers. However, customers may possess numerous debt obligations that credit bureaus uncover, which can help to measure their repayment ability. This study proposed a comparative methodology that leverages five machine learning algorithms to predict personal loan defaults using debt-to-income ratio apart from the credit scoring models that prevail at banks. It analyzed the impact of debt payments on loan defaults and applied ensemble clustering to categorize customers’ risk levels based on their debt-to-income ratio. Experimental results indicated that ensemble clustering has enhanced the prediction power compared to conventional classification models to predict loan defaults.
item: Conference-Full-text
Occlusion resilient similar-colored separable food item instance segmentation
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Karannagoda, R; Perera, Y; Weiman, D; Fernando, S; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
The task of recognizing non-Western and non- Chinese food items as well as accurately segmenting food item instances is a seldom researched and challenging task in the field of Computer Vision. Food items such as Sri Lankan short eats snacks have high inter-class visual similarity, mainly in terms of color and the fact that food images are highly prone to occlusion or item overlap where a portion of an object is hidden from sight. Existing databases are few and synthetic and current systems do not handle food item occlusion. In this paper a novel Sri Lankan short eats food item instance segmentation and amodal completion approach is introduced as well as two novel datasets for Sri Lankan short eats instance segmentation and amodal instance segmentation. The proposed method shows model performance improvements up to 88.4% mAP in Instance Segmentation and up to 90% mIoU in Amodal Completion, as well as the advantage of real-time inference in less than 1.7 seconds per frame.
item: Conference-Full-text
Generating photographic face images from sketches: a study of gan-based approaches
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Kovarthanan, K; Kumarasinghe, KMSJ; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
Generative Adversarial Networks (GANs) have attracted a lot of attention in recent years due to their potential to advance various fields. The high generative quality of GANs has been harnessed for creating photographic facial portraits from sketches in the field of computer vision. Given the increasing importance of computer vision, the ability to transform handdrawn sketches into realistic facial images has emerged as a compelling area of research. This practical implication can contribute to diverse fields, including law enforcement, forensics, security, and expedited generation of authentic suspect photos in crime investigations. Despite the inherent lack of specific information in sketch images, the training process necessitates meticulously crafted hand sketches to yield accurate and highquality results. This paper explores various approaches employed to address the challenges of translating facial sketches into photographic images, with a particular focus on GANs and their applications. The study aims to deliver a comprehensive analysis of state-of-the-art GAN-based methods for generating photographic faces from sketches. By offering a thorough overview of the strengths, methodologies, and advances in this field, this paper aims to pave the way for further advancements in the exciting area of sketch-to-photo face generation. Performance comparisons have been conducted among the different approaches in generating facial images from hand-drawn sketches, showcasing the effectiveness of several GAN architectures, each with a unique set of benefits and drawbacks.
item: Conference-Full-text
Green insight: a novel approach to detecting and classifying macro nutrient deficiencies in paddy leaves.
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Rathnayake, DMGD; Kumarasinghe, KMSJ; Rajapaksha, RMIK; Katuwawala, NKAC; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
Macro nutrient deficiency in paddy leaves is a critical concern in agriculture that impacts crop yield, food security, and sustainable farming. Addressing nutrient deficiencies in paddy plants is vital for ensuring these concerns. This research focuses on automating the detection and classification of common macro-nutrient deficiencies, specifically Nitrogen (N), Phosphorus (P), and Potassium (K). Utilizing image processing techniques, the study identifies distinct color patterns associated with each deficiency, providing a non-invasive and efficient approach. The analysis involves pixel ratio calculations within defined HSV color ranges and threshold values. A modular workflow encompasses preprocessing, horizontal partitioning, pixel ratio computation, and deficiency classification. The innovative methodology we introduced demonstrates promising outcomes, achieving a 96% accuracy rate in identifying nitrogen deficiency, along with 90% accuracy for phosphorus deficiency and 92% accuracy for potassium deficiency detection. While the methodology showcases promise, certain limitations, such as the requirement for leaf symmetry and single-deficiency identification, are recognized. These findings lay the groundwork for more accurate and automated nutrient deficiency detection, and the future work aims to address the identified limitations and generalize the solution for broader applications in real-world agricultural settings.
item: Conference-Full-text
Riceguardnet: custom cnns for precise bacterial and fungal infection classification
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Katuwawala, NKAC; Kumarasinghe, KMSJ; Rajapaksha, RMIK; Rathnayaka, DMGD; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
Rice cultivation is a vital component of many nations’ agricultural landscapes, often relying on traditional knowledge passed down through generations. However, disease identification in rice crops presents challenges, as many diseases are difficult to discern through visual inspection alone. This leads to delayed or inaccurate diagnoses, placing entire plantations at risk and discouraging new entrants to the field. This research addresses the pressing issue of timely and accurate disease identification in rice plants, focusing on three common diseases: Bacterial Leaf Blight, Brown Spot, and Leaf Smut, which are caused by bacteria and fungi. These diseases can proliferate rapidly, making early detection crucial. A custom Convolutional Neural Network (CNN) model was developed and trained using a dataset comprising 16,000 images, with 4,000 images for each disease and a healthy class. The model achieved an impressive accuracy of 99.87% on the test dataset, demonstrating its effectiveness in disease classification. This innovative approach provides a solution to the challenges faced by rice farmers, enabling quick and accurate disease identification. The research findings hold significant promise for improving rice cultivation practices, reducing the risk of crop loss, and encouraging new entrants into the field of rice farming.
item: Conference-Full-text
Cross-vit: cross-attention vision transformer for image duplicate detection
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Chandrasiri, MDN; Talagala, PD; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
Duplicate detection in image databases has immense significance across diverse domains. Its utility transcends specific applications, adapting seamlessly to a range of use cases, either as a standalone process or an integrated component within broader workflows. This study explores cutting-edge vision transformer architecture to revolutionize feature extraction in the context of duplicate image identification. Our proposed framework combines the conventional transformer architecture with a groundbreaking cross-attention layer developed specifically for this study. This unique cross-attention transformer processes pairs of images as input, enabling intricate cross-attention operations that delve into the interconnections and relationships between the distinct features in the two images. Through meticulous iterations of Cross-ViT, we assess the ranking capabilities of each version, highlighting the vital role played by the integrated cross-attention layer between transformer blocks. Our research culminates in recommending a final optimal model that capitalizes on the synergies between higher-dimensional hidden embeddings and mid-size ViT variations, thereby optimizing image pair ranking. In conclusion, this study unveils the immense potential of the vision transformer and its novel cross-attention layer in the domain of duplicate image detection. The performance of the proposed framework was assessed through a comprehensive comparative evaluation against baseline CNN models using various benchmark datasets. This evaluation further underscores the transformative power of our approach. Notably, our innovation in this study lies not in the introduction of new feature extraction methods but in the introduction of a novel cross-attention layer between transformer blocks grounded in the scaled dot-product attention mechanism.
item: Conference-Full-text
Resume content scoring and improvement suggestions using nlp and rule-based techniques
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Weerasinghe, RL; Perera, NN; Warusawithana, SP; Hindakaraldeniya, TM; Ganegoda, GU; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
Having a proper resume is very important for undergraduates or fresh graduates to find their dream job. But most of them find it difficult to prepare their resume properly by themselves. It often needs a third party to review the resume to identify missing parts and content improvements of the resume because most of the time candidates make some mistakes. When it comes to resume review systems, most of the systems are based on the recruiter perspective which does not provide any insights for the candidate to improve their resumes. Hence, it is helpful if a proper resume content reviewer is there for candidates to analyze their resumes. This study is focused on developing a model to resume content scoring and suggest missing content based on NLP and rule-based techniques. Two separate approaches were developed and tested for the proposed system and then the comparison of those approaches were carried out through this study.
item: Conference-Full-text
Resbot: a bi-lingual restaurant booking conversational artificial intelligence
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Fernando, LKD; Ganegoda, GU; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
AI-powered chatbots in the service industry enhance customer service and enable data-driven decisions. However, in a conversion, the user's input may not always align closely with the training examples, causing even advanced Natural Language Understanding pipelines to occasionally misinterpret the intent behind user utterances, leading to potential conversational missteps. This paper introduces ResBot, an innovative bi-lingual chatbot (Sinhala and English) with a novel hybrid intent classification mechanism. This approach emphasizes the importance of generalization in intent recognition beyond training data using a language model. Furthermore, by automating reservations through its chat interface, the chatbot transforms customer experience and optimizes restaurant operations in an increasingly digital landscape.
item: Conference-Full-text
Explainable ai techniques for deep convolutional neural network based plant disease identification
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Kiriella, S; Fernando, S; Sumathipala, S; Udayakumara, EPN; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
Deep learning-based computer vision has shown improved performance in image classification tasks. Due to the complexities of these models, they have been referred as opaque models. As a result, users need justifications for predictions to enhance trust. Thus, Explainable Artificial Intelligence (XAI) provides various techniques to explain predictions. Explanations play a vital role in practical application, to apply the exact treatment for a plant disease. However, application of XAI techniques in plant disease identification is not popular. This paper discusses the key concerns and taxonomies available in XAI and summarizes the recent developments. Also, it develops a tomato disease classification model and uses different XAI techniques to validate model predictions. It includes a comparative analysis of XAI techniques and discusses the limitations and usefulness of the techniques in plant disease symptom localization.
item: Conference-Full-text
Improved particle swarm optimization for optimizing the deep convolutional neural network
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Atugoda, AWCK; Fernando, S; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
In recent years, Deep Neural Networks (DNN) have been employed in different types of fields for recognizing, classifying, detecting and sorting, etc. Thus, optimizing the DNN is very essential to obtain a potential solution with high accuracy. Neural network(NN) can be optimized by optimizing the weight values of the network. Many studies have been done utilizing conventional optimization techniques such as Stochastic Gradient Descent(SGD), Adam, Ada Delta, and so on. Employing traditional optimization approaches in optimizing the deep neural network, on the other hand, results in poor performance due to trapping at local extremes and premature convergence. As a result, researchers looked into Swarm Intelligence(SI) optimization algorithms, which are fast and robust global optimization methods that have gained a lot of attention due to their capability to deal with complicated optimization problems. Among different types of SI algorithms, Particle Swarm Optimization (PSO) is mostly used in NN optimization as it has a few parameters to be tuned, and no derivative for simplification. However, recent studies have shown that the standard PSO is not the best tool for tackling all engineering problems since it is slow in some contexts, such as biomedical engineering and building construction, and converges to local optima. Therefore, improving the PSO algorithm is critical for obtaining a feasible solution to NN optimization problems. Hence, the main goal of this study is to make advanced enhancements to the PSO algorithm to optimize DNN while addressing several concerns, such as minimizing the computational cost or Graphical Processing Unit (GPU) dependency and having large input data in Deep Convolutional Neural Network (DCNN) training.
item: Conference-Full-text
Learning application for educational and skills development of primary children
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Malshika, MDJ; Wijeratne, NS; Kavishka, PKP; Chathurika, B; Karunathilaka, S; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
This research uses interactive web-based learning tools to improve primary school children's language skills, mathematical competency, and critical thinking ability. The study attempts to bridge the gap between linguistic and numerical literacy while also including critical thinking aspects into the learning process, resulting in a full and interesting educational experience. The importance of this study stems from its dedication to comprehensive growth through the integration of technology and education. The technique integrates auditory, visual, and interactive components to make language and mathematical skill development successful and pleasant by providing a dynamic platform that accommodates varied learning styles. The project's methodology entails rigorous data gathering, the building of strong neural network models for linguistic and arithmetic skills, and the use of facial expression detection technologies to test critical thinking skills. Through various data sources, these models are taught to recognize and improve handwriting, numbers, and mathematical problemsolving. The study yields favorable results in all three categories. Significant increases in letter and numerical recognition, vocabulary enhancement, and mathematical competency have been demonstrated via interactive games aimed to promote language and mathematics ability. Furthermore, the use of facial expression detection technologies in educational games analyses and improves the critical thinking skills of primary school children. Finally, this study pioneers a strategy that demonstrates the potential of interactive web-based learning apps to improve linguistic, mathematics, and critical thinking skills. The findings reflect a huge step forward towards a more integrated and effective learning environment, with technology aiding an all-encompassing education. The research emphasizes the need to integrate technology and education to provide kids with language, numeracy, and critical thinking abilities that will prepare them for future problems.
item: Conference-Full-text
Classification of fungi images using different convolutional neural networks
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Nawarathne, UMMPK; Kumari, HMNS; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
Fungi offer vital solutions to humanity through roles in medicine, agriculture, and ecological balance while presenting potential threats. They have yielded antibiotics, food fermentation, and nutrient recycling however, fungal infections, crop diseases, and spoilage highlight their dark side. Therefore, it is important to identify fungi to harness their potential benefits and mitigate threats. Offering quick and accurate identification through image classification improves the aforementioned features. Therefore, this study classified images of five types of fungi using convolutional neural networks (CNN). Initially, dataset distribution was observed, and it was identified that there was a class imbalance in the dataset. To address this issue, data augmentation technique was used. Several preprocessing techniques were also applied to understand the model training behavior with their application. Then the images were rescaled into six different resolution combinations such as original images, low-resolution images, high-resolution images, a mix of original and low-resolution images, a mix of original and high-resolution images, and a mix of low and high-resolution images. Then these data were trained using 13 pre-trained CNN models such as Xception, VGG16, VGG19, InceptionResNetV2, ResNet152, EfficientNetB6, EfficientNetB7, ConvNeXtTiny, ConvNeXtSmall, ConvNeXtBase, ConvNeXtLarge, ConvNeXtXLarge, BigTransfer (BiT). To evaluate these models, accuracy, macro average precision, macro average recall, macro average f1- score, and loss learning curve assessment were used. According to the results, the BiT model preprocessed with normalization, which used a mix of original and high-resolution images, performed the best, producing a model accuracy of 87.32% with optimal precision, recall, and f1-score. The loss learning curve of the BiT model also depicted a low overfitting aspect proving the model’s optimal behavior. Therefore, it was concluded that the BiT model with the mix of original and high-resolution data can be used to detect fungi efficiently.
item: Conference-Full-text
Enhancing ddos attack detection via blending ensemble learning
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Amalraj, CRJ; Madhusankha, PGG; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
This research focuses on identifying DDoS attacks using an ensemble learning approach that incorporates blending techniques. We developed an innovative methodology by selecting the 21 most significant features from the CIC-DDoS2019 dataset. To improve classification accuracy, we used a two-layer blending ensemble technique. In the first layer, we combined Decision Tree, Logistic Regression, and KNN classifiers, while the second layer used a Random Forest classifier. The model achieved exceptional results, with a 99.94% accuracy score and a 97.35% F1 score for detecting DDoS attacks accurately. We also created a user-friendly web portal to make the model accessible for individuals in network security, regardless of their technical expertise. This approach advances DDoS attack detection and enhances usability for users in the field of network security.
item: Conference-Full-text
Alzheimer’s disease prediction using clinical data approach
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Perera, LRD; Ganegoda, GU; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
Alzheimer's Disease (AD) is a progressive neurodegenerative condition that profoundly affects cognition and memory. Due to the absence of curative treatments, early detection and prediction are crucial for effective intervention. This study employs machine learning and clinical data from Alzheimer's Disease Neuroimaging Initiative (ADNI) to predict AD onset. Data preprocessing ensures quality through variable selection and feature extraction. Diverse machine learning algorithms, including Naive Bayes, logistic regression, SVM-Linear, random forest, Gradient Boosting, and Decision Trees, are evaluated for prediction accuracy. The model resulted with random forest classifier together with filter method yields the highest AUC. The study highlights important analysis using Random Forest and Decision Trees, revealing significant variables including cognitive tests, clinical scales, demographics, brain-related metrics, and key biomarkers. By enhancing predictive capabilities, this research contributes to advancing Alzheimer's disease diagnosis and intervention strategies.
item: Conference-Full-text
Alzheimer’s disease detection using blood gene expression data
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Yasodya, GDS; Ganegoda, GU; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
Alzheimer's disease is the most prevalent form of dementia with no established cure. Extensive research aims to comprehend its underlying mechanisms. Genetic insights are sought through gene expression data analysis, leveraging computational and statistical techniques to identify risk-associated genes. This study focuses on accurate AD detection using blood gene expression data. Four feature classification methods—TFrelated genes, Hub genes, CFG, and VAE are employed to identify crucial AD-related genes. Five classification approaches—RF, SVM, LR, L1-LR, and DNN—are used, evaluated by AUC. The VAE + LR model yields the highest AUC (0.76). The study identifies 100 influential AD-associated genes where data is sourced from Alzheimer's Disease Neuroimaging Initiative (ADNI). Findings hold promise for advancing early diagnosis and treatment, enhancing AD patients' quality of life.
item: Conference-Full-text
Performance improvement of proxy server cache management using web usage mining
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2023-12-07) Leenas, T; Caldera, HA; Piyatilake, ITS; Thalagala, PD; Ganegoda, GU; Thanuja, ALARR; Dharmarathna, P
In these modern industries, all sectors are transitioning from manual to web-oriented applications. Thus, the number of Internet users are increasing drastically. Therefore, there is a substantial traffic, which increases the demand on the server and server response latency to obtain web objects. The proxy server caching mechanism is one of the approaches to enhance the performance of accessing web objects via the Internet. Since the cache is typically limited in size, a replacement strategy is required to decide which cached web object should be eliminated to allow spaces for fresh web items. Proxy servers make use of various cache replacement strategies such as Least Recency Used (LRU), Least Frequently Used (LFU), and SIZE. The web objects in the proxy cache are influenced by variables like recency, frequency, fetching time, and size. The traditional caching policies decide only one factor at a time, unpopular web objects are wasted in the cache memory (cache pollution) and the performance of the proxy cache decreases. To increase its performance, we propose using the proxy server log file to identify individual users and their sessions as well as categorize the web objects into three groups: high priority, average priority, and low priority web objects. The prepared log file is used to train the classifiers. Future requests are classified as high, average, or low objects using the classifiers, and it is then chosen whether to store them in the proxy cache or not. The objective of this research is to enhance the proxy caching mechanism by implementing the techniques mentioned above. We compared the performance of the suggested approach with traditional caching policies using a trace-driven simulation method. Two performance metrics, Hit Ratio (HR), and Byte Bit Ratio (BHR), were used for our investigation. Our experimental findings demonstrate that the suggested approach outperforms traditional caching policies.

Browse

Recent Submissions