Pub Date : 2023-10-01DOI: 10.7494/csci.2023.24.3.4367
Ha-Bang Ban, Hong-Phuong Nguyen, Dang-Hai Pham
The School Bus-Driver Problem with Resource Constraints (SBDP-RC) is an optimization problem with many practical applications. In the problem, the number of vehicles is prepared to pick a number of pupils, in which the total resource of all vehicles is less than a predefined value. The aim is to find a tour minimizing the sum of pupils’ waiting times. The problem is NP-hard in the general case. In many cases, reaching a feasible solution becomes an NP-hard problem. To solve the large-sized problem, a metaheuristic approach is a suitable approach. The first phase creates an initial solution by the construction heuristic based on Insertion Heuristic. After that, the post phase improves the solution by the General Variable Neighborhood Search (GVNS) with Random Neighborhood Search combined with Shaking Technique. The hybridization ensures the balance between exploitation and exploration. Therefore, the proposed algorithm can escape from local optimal solutions. The proposed metaheuristic algorithm is tested on a benchmark to show the efficiency of the algorithm. The results show that the algorithm receives good feasible solutions fast. Additionally, in many cases, better solutions can be found in comparison with the previous metaheuristic algorithms.
{"title":"Hybrid Variable Neighborhood Search for Solving School Bus-Driver Problem with Resource Constraints","authors":"Ha-Bang Ban, Hong-Phuong Nguyen, Dang-Hai Pham","doi":"10.7494/csci.2023.24.3.4367","DOIUrl":"https://doi.org/10.7494/csci.2023.24.3.4367","url":null,"abstract":"The School Bus-Driver Problem with Resource Constraints (SBDP-RC) is an optimization problem with many practical applications. In the problem, the number of vehicles is prepared to pick a number of pupils, in which the total resource of all vehicles is less than a predefined value. The aim is to find a tour minimizing the sum of pupils’ waiting times. The problem is NP-hard in the general case. In many cases, reaching a feasible solution becomes an NP-hard problem. To solve the large-sized problem, a metaheuristic approach is a suitable approach. The first phase creates an initial solution by the construction heuristic based on Insertion Heuristic. After that, the post phase improves the solution by the General Variable Neighborhood Search (GVNS) with Random Neighborhood Search combined with Shaking Technique. The hybridization ensures the balance between exploitation and exploration. Therefore, the proposed algorithm can escape from local optimal solutions. The proposed metaheuristic algorithm is tested on a benchmark to show the efficiency of the algorithm. The results show that the algorithm receives good feasible solutions fast. Additionally, in many cases, better solutions can be found in comparison with the previous metaheuristic algorithms.","PeriodicalId":41917,"journal":{"name":"Computer Science-AGH","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135406613","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-10-01DOI: 10.7494/csci.2023.24.3.5116
Anuradha Ariyaratne, I M T P K Ilankoon, U Samarasinghe, R M Silva
Cluster analysis can be defined as applying clustering algorithms with the goal of finding hidden patterns or groupings in a dataset. Different clustering methods provide different solutions for the same dataset. Traditional clustering algorithms are popular, but handling big data sets is beyond the ability of such methods. We propose three big data clustering methods, based on the Firefly Algorithm (FA). Three different fitness functions were defined on FA using inter cluster distance, intra cluster distance, silhouette value and Calinski-Harabasz Index. The algorithms find the most appropriate cluster centers for a given data set. The algorithms were tested with four popular synthetic data sets and later applied on two badminton data sets to identify different playing styles of players based on physical characteristics. The results specify that the firefly algorithm could generate better clustering results with high accuracy. The algorithms cluster the players to find the most suitable playing strategy for a given player where expert knowledge is needed in labeling the clusters. Comparisons with a PSO based clustering algorithm (APSO) and traditional algorithms point out that the proposed firefly variants work similarly as the APSO method and surpass the performance of traditional algorithms.
{"title":"Finding Playing Styles of Badminton Players Using Firefly Algorithm Based Clustering Algorithms","authors":"Anuradha Ariyaratne, I M T P K Ilankoon, U Samarasinghe, R M Silva","doi":"10.7494/csci.2023.24.3.5116","DOIUrl":"https://doi.org/10.7494/csci.2023.24.3.5116","url":null,"abstract":"Cluster analysis can be defined as applying clustering algorithms with the goal of finding hidden patterns or groupings in a dataset. Different clustering methods provide different solutions for the same dataset. Traditional clustering algorithms are popular, but handling big data sets is beyond the ability of such methods. We propose three big data clustering methods, based on the Firefly Algorithm (FA). Three different fitness functions were defined on FA using inter cluster distance, intra cluster distance, silhouette value and Calinski-Harabasz Index. The algorithms find the most appropriate cluster centers for a given data set. The algorithms were tested with four popular synthetic data sets and later applied on two badminton data sets to identify different playing styles of players based on physical characteristics. The results specify that the firefly algorithm could generate better clustering results with high accuracy. The algorithms cluster the players to find the most suitable playing strategy for a given player where expert knowledge is needed in labeling the clusters. Comparisons with a PSO based clustering algorithm (APSO) and traditional algorithms point out that the proposed firefly variants work similarly as the APSO method and surpass the performance of traditional algorithms.","PeriodicalId":41917,"journal":{"name":"Computer Science-AGH","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135406941","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-10-01DOI: 10.7494/csci.2023.24.3.4831
Rafal Mucha, Bartosz Balis, Costin Grigoras, Jacek Kitowski
Database replication is a way to improve system throughput or achieve high availability. In most cases, using an active-active replica architecture is efficient and easy to deploy. Such a system has CP properties (from the CAP theorem: Consistency, Availability and network Partition tolerance). Creating an AP (available and partition tolerant) system requires using multi-primary replication. This approach, because of many difficulties in implementation, is not widely used. However, deployment of CCDB (experiment conditions and calibration database) needs to be an AP system in two locations. This necessity became an inspiration to examine the state-of-the-art in this field and to test the available solutions. The tests performed evaluate the performance of the chosen replication tools: Bucardo and EDB Replication Server. They show that the tested tools can be successfully used for continuous synchronization of two independent database instances.
{"title":"Database Replication for Disconnected Operations with Quasi Real-Time Synchronization","authors":"Rafal Mucha, Bartosz Balis, Costin Grigoras, Jacek Kitowski","doi":"10.7494/csci.2023.24.3.4831","DOIUrl":"https://doi.org/10.7494/csci.2023.24.3.4831","url":null,"abstract":"Database replication is a way to improve system throughput or achieve high availability. In most cases, using an active-active replica architecture is efficient and easy to deploy. Such a system has CP properties (from the CAP theorem: Consistency, Availability and network Partition tolerance). Creating an AP (available and partition tolerant) system requires using multi-primary replication. This approach, because of many difficulties in implementation, is not widely used. However, deployment of CCDB (experiment conditions and calibration database) needs to be an AP system in two locations. This necessity became an inspiration to examine the state-of-the-art in this field and to test the available solutions. The tests performed evaluate the performance of the chosen replication tools: Bucardo and EDB Replication Server. They show that the tested tools can be successfully used for continuous synchronization of two independent database instances.","PeriodicalId":41917,"journal":{"name":"Computer Science-AGH","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135406611","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Deep learning models form one of the most powerful machine learning models for the extraction of important features. Most of the designs of deep neural models, i.e., the initialization of parameters, are still manually tuned. Hence, obtaining a model with high performance is exceedingly time-consuming and occasionally impossible. Optimizing the parameters of the deep networks, therefore, requires improved optimization algorithms with high convergence rates. The single objective-based optimization methods generally used are mostly time-consuming and do not guarantee optimum performance in allcases. Mathematical optimization problems containing multiple objective functions that must be optimized simultaneously fall under the category of multi-objective optimization sometimes referred to as Pareto optimization. Multi-objective optimization problems form one of the alternatives yet useful options for parameter optimization. However, this domain is a bit less explored. In this survey, we focus on exploring the effectiveness of multi-objective optimization strategies for parameter optimization in conjunction with deep neural networks. The case studies used in this study focus on how the twomethods are combined to provide valuable insights into the generation of predictions and analysis in multiple applications.
{"title":"A Survey on Multi-Objective Based Parameter Optimization for Deep Learning","authors":"Mrittika Chakraborty, Wreetbhas Pal, Sanghamitra Bandyopadhyay, Ujjwal Maulik","doi":"10.7494/csci.2023.24.3.5479","DOIUrl":"https://doi.org/10.7494/csci.2023.24.3.5479","url":null,"abstract":"Deep learning models form one of the most powerful machine learning models for the extraction of important features. Most of the designs of deep neural models, i.e., the initialization of parameters, are still manually tuned. Hence, obtaining a model with high performance is exceedingly time-consuming and occasionally impossible. Optimizing the parameters of the deep networks, therefore, requires improved optimization algorithms with high convergence rates. The single objective-based optimization methods generally used are mostly time-consuming and do not guarantee optimum performance in allcases. Mathematical optimization problems containing multiple objective functions that must be optimized simultaneously fall under the category of multi-objective optimization sometimes referred to as Pareto optimization. Multi-objective optimization problems form one of the alternatives yet useful options for parameter optimization. However, this domain is a bit less explored. In this survey, we focus on exploring the effectiveness of multi-objective optimization strategies for parameter optimization in conjunction with deep neural networks. The case studies used in this study focus on how the twomethods are combined to provide valuable insights into the generation of predictions and analysis in multiple applications.","PeriodicalId":41917,"journal":{"name":"Computer Science-AGH","volume":"150 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135406617","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-10-01DOI: 10.7494/csci.2023.24.3.4844
N. I. Md. Ashafuddula, Rafiqul Islam
Melanoma skin cancer is one of the most dangerous and life-threatening cancer. Exposure to ultraviolet rays may damage the skin cell's DNA, which causes melanoma skin cancer. However, it is difficult to detect and classify melanoma and nevus mole at the immature stages. In this work, an automatic deep learning system is developed based on the intensity value estimation with a convolutional neural network model (CNN) to detect and classify melanoma and nevus mole more accurately. Since intensity levels are the most distinctive features for object or region of interest identification, the high-intensity pixel values are selected from the extracted lesion images. Incorporating those high-intensity features into the CNN improves the overall performance than the state-of-the-art methods for detecting melanoma skin cancer. To evaluate the system, we used 5-fold cross-validation. Experimental results show that a superior percentage of accuracy (92.58%), Sensitivity (93.76%), Specificity (91.56%), and Precision (90.68%) are achieved.
{"title":"Melanoma Skin Cancer and Nevus Mole Classification using Intensity Value Estimation with Convolutional Neural Network","authors":"N. I. Md. Ashafuddula, Rafiqul Islam","doi":"10.7494/csci.2023.24.3.4844","DOIUrl":"https://doi.org/10.7494/csci.2023.24.3.4844","url":null,"abstract":"Melanoma skin cancer is one of the most dangerous and life-threatening cancer. Exposure to ultraviolet rays may damage the skin cell's DNA, which causes melanoma skin cancer. However, it is difficult to detect and classify melanoma and nevus mole at the immature stages. In this work, an automatic deep learning system is developed based on the intensity value estimation with a convolutional neural network model (CNN) to detect and classify melanoma and nevus mole more accurately. Since intensity levels are the most distinctive features for object or region of interest identification, the high-intensity pixel values are selected from the extracted lesion images. Incorporating those high-intensity features into the CNN improves the overall performance than the state-of-the-art methods for detecting melanoma skin cancer. To evaluate the system, we used 5-fold cross-validation. Experimental results show that a superior percentage of accuracy (92.58%), Sensitivity (93.76%), Specificity (91.56%), and Precision (90.68%) are achieved.","PeriodicalId":41917,"journal":{"name":"Computer Science-AGH","volume":"63 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135406935","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper presents a hybrid meta-heuristic algorithm using Grey Wolf optimization (GWO) and JAYA algorithm for data clustering. The idea is use exploitative capability of JAYA algorithm in the explorative phase of GWO to form compact clusters. Here, instead of using one best and one worst solution for generating offspring, three best wolfs and three worst omega wolfs of the population are used. So, the best wolfs and worst omega wolfs assist in moving the new solutions towards the best solutions and simultaneously helps in staying away from the worst solutions. This enhances the chances of reaching the near optimal solutions. The superiority of the proposed method is compared with five promising algorithms, namely GWO, Sine-Cosine Algorithm (SCA), Particle Swarm Optimization (PSO), JAYA and K-means algorithms. The result obtained from the Duncan’s multiple range test and Nemenyi hypothesis based statistical test confirms the superiority and robustness of our proposed method.
{"title":"A Nature Inspired Hybrid Partitional Clustering Method Based on Grey Wolf Optimization and JAYA Algorithm","authors":"GYANARANJAN SHIAL, Sabita Sahoo, Sibarama Panigrahi","doi":"10.7494/csci.2023.24.3.4962","DOIUrl":"https://doi.org/10.7494/csci.2023.24.3.4962","url":null,"abstract":"This paper presents a hybrid meta-heuristic algorithm using Grey Wolf optimization (GWO) and JAYA algorithm for data clustering. The idea is use exploitative capability of JAYA algorithm in the explorative phase of GWO to form compact clusters. Here, instead of using one best and one worst solution for generating offspring, three best wolfs and three worst omega wolfs of the population are used. So, the best wolfs and worst omega wolfs assist in moving the new solutions towards the best solutions and simultaneously helps in staying away from the worst solutions. This enhances the chances of reaching the near optimal solutions. The superiority of the proposed method is compared with five promising algorithms, namely GWO, Sine-Cosine Algorithm (SCA), Particle Swarm Optimization (PSO), JAYA and K-means algorithms. The result obtained from the Duncan’s multiple range test and Nemenyi hypothesis based statistical test confirms the superiority and robustness of our proposed method.","PeriodicalId":41917,"journal":{"name":"Computer Science-AGH","volume":"67 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135406609","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The dominating set problem in graph theory is an NP-complete problem for an arbitrary graph. There are many approximation-based studies in the literature to solve the dominating set problems for a given graph. Some of them are exact algorithms with exponential time complexities and some of them are based on approximation without robustness with respect to obtained solutions. In this study, the Malatya centrality value was used and a new Malatya centrality value was defined to solve the dominating set problem for a given graph. The improved algorithms have polynomial time and space complexities.
{"title":"Calculating the Centrality Values According to the Strengths of Entities Relative to their Neighbours and Designing a New Algorithm for the Solution of the Minimal Dominating Set Problem","authors":"Şeyda Karci, F. Okumuş, A. Karcı","doi":"10.53070/bbd.1295038","DOIUrl":"https://doi.org/10.53070/bbd.1295038","url":null,"abstract":"The dominating set problem in graph theory is an NP-complete problem for an arbitrary graph. There are many approximation-based studies in the literature to solve the dominating set problems for a given graph. Some of them are exact algorithms with exponential time complexities and some of them are based on approximation without robustness with respect to obtained solutions. In this study, the Malatya centrality value was used and a new Malatya centrality value was defined to solve the dominating set problem for a given graph. The improved algorithms have polynomial time and space complexities.","PeriodicalId":41917,"journal":{"name":"Computer Science-AGH","volume":"1 1","pages":""},"PeriodicalIF":0.5,"publicationDate":"2023-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42428858","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Email, which stands for electronic mail, is a form of digital communication between two or more individuals. These technological instruments that facilitate communication can have a positive and negative impact on our lives due to junk e-mails, widely known as spam mail. These spam messages, which are typically delivered for commercial purposes by organizations/individuals for indirect or direct benefits, not only distract people but also consume a significant amount of system resources such as processing power, memory, and network bandwidth. In this study, a method based on LBP (Local Binary Patterns) feature extraction and statistical pooling is proposed to classify spam or raw (non-spam) images. Two datasets are used to test the proposed method. The ISH dataset is widely used in the literature and contains 1738 images. In addition to this dataset, the dataset our collect consists of 1015 images in total. Feature extraction was performed on these images. Obtained features were classified by SVM (Support Vector Machine) algorithm. In the proposed method, 98.56% and 79.01% accuracy were calculated for the ISH dataset and our collected dataset, respectively. The results obtained were compared with the studies in the literature.
{"title":"LBP Özellik Çıkarma ve İstatistiksel Havuzlama Tabanlı Görüntü Spam Tespit Modeli","authors":"Aytaç Kaşoğlu, Orhan Yaman","doi":"10.53070/bbd.1268221","DOIUrl":"https://doi.org/10.53070/bbd.1268221","url":null,"abstract":"Email, which stands for electronic mail, is a form of digital communication between two or more individuals. These technological instruments that facilitate communication can have a positive and negative impact on our lives due to junk e-mails, widely known as spam mail. These spam messages, which are typically delivered for commercial purposes by organizations/individuals for indirect or direct benefits, not only distract people but also consume a significant amount of system resources such as processing power, memory, and network bandwidth. In this study, a method based on LBP (Local Binary Patterns) feature extraction and statistical pooling is proposed to classify spam or raw (non-spam) images. Two datasets are used to test the proposed method. The ISH dataset is widely used in the literature and contains 1738 images. In addition to this dataset, the dataset our collect consists of 1015 images in total. Feature extraction was performed on these images. Obtained features were classified by SVM (Support Vector Machine) algorithm. In the proposed method, 98.56% and 79.01% accuracy were calculated for the ISH dataset and our collected dataset, respectively. The results obtained were compared with the studies in the literature.","PeriodicalId":41917,"journal":{"name":"Computer Science-AGH","volume":" ","pages":""},"PeriodicalIF":0.5,"publicationDate":"2023-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"48272692","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Meta-heuristics are high-level approaches developed to discover a heuristic that provides a reasonable solution to many varieties of optimization problems. The classification problems contain a sort of optimization problem. Simply, the objective herein is to reduce the number of misclassified instances. In this paper, the question of whether meta-heuristic methods can be used to construct linear models or not is answered. To this end, Particle Swarm Optimization (PSO) has been engaged to address linear classification problems. The Particle Swarm Classifier (PSC) with a certain objective function has been compared with Support Vector Machine (SVM), Perceptron Learning Rule (PLR), and Logistic Regression (LR) applied to fifteen data sets. The experimental results point out that PSC can compete with the other classifiers, and it turns out to be superior to other classifiers for some binary classification problems. Furthermore, the average classification accuracies of PSC, SVM, LR, and PLR are 80.8%, 80.6%, 80.9%, and 57.7%, respectively. In order to enhance the classification performance of PSC, more advanced objective functions can be developed. Further, the classification accuracy can be boosted more by constructing tighter constraints via another meta-heuristic.
{"title":"Parçacık Sürü Optimizasyonu Yoluyla Geliştirilen Doğrusal Bir Sınıflandırıcının Analizi","authors":"F. Aydin","doi":"10.53070/bbd.1259377","DOIUrl":"https://doi.org/10.53070/bbd.1259377","url":null,"abstract":"Meta-heuristics are high-level approaches developed to discover a heuristic that provides a reasonable solution to many varieties of optimization problems. The classification problems contain a sort of optimization problem. Simply, the objective herein is to reduce the number of misclassified instances. In this paper, the question of whether meta-heuristic methods can be used to construct linear models or not is answered. To this end, Particle Swarm Optimization (PSO) has been engaged to address linear classification problems. The Particle Swarm Classifier (PSC) with a certain objective function has been compared with Support Vector Machine (SVM), Perceptron Learning Rule (PLR), and Logistic Regression (LR) applied to fifteen data sets. The experimental results point out that PSC can compete with the other classifiers, and it turns out to be superior to other classifiers for some binary classification problems. Furthermore, the average classification accuracies of PSC, SVM, LR, and PLR are 80.8%, 80.6%, 80.9%, and 57.7%, respectively. In order to enhance the classification performance of PSC, more advanced objective functions can be developed. Further, the classification accuracy can be boosted more by constructing tighter constraints via another meta-heuristic.","PeriodicalId":41917,"journal":{"name":"Computer Science-AGH","volume":" ","pages":""},"PeriodicalIF":0.5,"publicationDate":"2023-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"46225999","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-03-10DOI: 10.7494/csci.2023.24.2.4761
Anmol Bansal, Arjun Choudhry, Anubhav Sharma, Seba Susan
Covid-19 has spread across the world and many different vaccines have been developed to counter its surge. To identify the correct sentiments associated with the vaccines from social media posts, this paper aims to fine-tune pre-trained transformer models on tweets associated with different Covid vaccines, specifically RoBERTa, XLNet and BERT which are recently introduced state-of-the-art bi-directional transformer models, and domain-specific transformer models BERTweet and CT-BERT that are pre-trained on Covid-19 tweets. We further explore the option of data augmentation by text oversampling using LMOTE to improve the accuracies of these models, specifically, for small sample datasets where there is an imbalanced class distribution among the positive, negative and neutral sentiment classes. Our results summarize our findings on the suitability of text oversampling for imbalanced, small sample datasets that are used to fine-tune state-of-the-art pre-trained transformer models, and the utility of having domain-specific transformer models for the classification task.
{"title":"ADAPTATION OF DOMAIN-SPECIFIC TRANSFORMER MODELS WITH TEXT OVERSAMPLING FOR SENTIMENT ANALYSIS OF SOCIAL MEDIA POSTS ON COVID-19 VACCINE","authors":"Anmol Bansal, Arjun Choudhry, Anubhav Sharma, Seba Susan","doi":"10.7494/csci.2023.24.2.4761","DOIUrl":"https://doi.org/10.7494/csci.2023.24.2.4761","url":null,"abstract":"Covid-19 has spread across the world and many different vaccines have been developed to counter its surge. To identify the correct sentiments associated with the vaccines from social media posts, this paper aims to fine-tune pre-trained transformer models on tweets associated with different Covid vaccines, specifically RoBERTa, XLNet and BERT which are recently introduced state-of-the-art bi-directional transformer models, and domain-specific transformer models BERTweet and CT-BERT that are pre-trained on Covid-19 tweets. We further explore the option of data augmentation by text oversampling using LMOTE to improve the accuracies of these models, specifically, for small sample datasets where there is an imbalanced class distribution among the positive, negative and neutral sentiment classes. Our results summarize our findings on the suitability of text oversampling for imbalanced, small sample datasets that are used to fine-tune state-of-the-art pre-trained transformer models, and the utility of having domain-specific transformer models for the classification task.","PeriodicalId":41917,"journal":{"name":"Computer Science-AGH","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"136096163","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：