Pub Date : 2021-12-18DOI: 10.1109/ICCIT54785.2021.9689853
M. Hossain, Mohammed Sowket Ali, Reshma Ahmed Swarna, M. Hasan, Nahida Habib, M. Rahman, M. Azad, Mohammad Motiur Rahman
A quantum feature map encodes classical data to the quantum state space by using a quantum circuit. The repetition of such a circuit during encoding is a customize value known as depth. Encoding data to quantum state is a must step for applying Quantum machine learning (QML) to classical data. Utilizing different feature map techniques by varying several depths, this research uses a kernel-based quantum support vector machine (QSVM) to classify several datasets. The fundamental aim of such activities is to check whether feature map techniques can make any sense to supervised QML concerning their depths and the outcomes analysis concludes that maximum accuracy of any supervised QML model is obtained due to the selection of an essential feature map approach with appropriate circuit depth. The results also present that time consumption of any feature map technique increases linearly with the increase of feature map circuit depth. However, the outcome of this research will help anyone to estimate the feature map technique and circuit depth when executing QML.
{"title":"Analyzing the effect of feature mapping techniques along with the circuit depth in quantum supervised learning by utilizing quantum support vector machine","authors":"M. Hossain, Mohammed Sowket Ali, Reshma Ahmed Swarna, M. Hasan, Nahida Habib, M. Rahman, M. Azad, Mohammad Motiur Rahman","doi":"10.1109/ICCIT54785.2021.9689853","DOIUrl":"https://doi.org/10.1109/ICCIT54785.2021.9689853","url":null,"abstract":"A quantum feature map encodes classical data to the quantum state space by using a quantum circuit. The repetition of such a circuit during encoding is a customize value known as depth. Encoding data to quantum state is a must step for applying Quantum machine learning (QML) to classical data. Utilizing different feature map techniques by varying several depths, this research uses a kernel-based quantum support vector machine (QSVM) to classify several datasets. The fundamental aim of such activities is to check whether feature map techniques can make any sense to supervised QML concerning their depths and the outcomes analysis concludes that maximum accuracy of any supervised QML model is obtained due to the selection of an essential feature map approach with appropriate circuit depth. The results also present that time consumption of any feature map technique increases linearly with the increase of feature map circuit depth. However, the outcome of this research will help anyone to estimate the feature map technique and circuit depth when executing QML.","PeriodicalId":166450,"journal":{"name":"2021 24th International Conference on Computer and Information Technology (ICCIT)","volume":"294 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132269773","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 : 2021-12-18DOI: 10.1109/ICCIT54785.2021.9689854
Dola Das, K. R. Alam, Y. Morimoto
Skyline query is well-known to find out the dominant objects from a large number of datasets. While multiple organizations want to analyze their combined dataset, skyline queries can assist in this regard. Maintaining privacy along with the data integrity of participating organizations’ datasets is important because their commercial success depends on the result of these queries. This paper proposes a new framework for the multi-party skyline query that encompasses both privacy and data integrity. To ensure the privacy of participants’ datasets, it adopts commutative encryptions by employing multiple independent entities. To support the data integrity, it combines encrypted unique tags (UTs) with the encrypted datasets of all participants. In addition, to retain the anonymity of participants’ encrypted data from anyone including authorities, it exploits the re-encryption. Although the proposed framework also practices homomorphic encryption, which usually sacrifices the data integrity, here due to the usage of UTs, it is maintained. This paper is a preliminary report of the proposed framework.
{"title":"A Framework for Multi-party Skyline Query Maintaining Privacy and Data Integrity","authors":"Dola Das, K. R. Alam, Y. Morimoto","doi":"10.1109/ICCIT54785.2021.9689854","DOIUrl":"https://doi.org/10.1109/ICCIT54785.2021.9689854","url":null,"abstract":"Skyline query is well-known to find out the dominant objects from a large number of datasets. While multiple organizations want to analyze their combined dataset, skyline queries can assist in this regard. Maintaining privacy along with the data integrity of participating organizations’ datasets is important because their commercial success depends on the result of these queries. This paper proposes a new framework for the multi-party skyline query that encompasses both privacy and data integrity. To ensure the privacy of participants’ datasets, it adopts commutative encryptions by employing multiple independent entities. To support the data integrity, it combines encrypted unique tags (UTs) with the encrypted datasets of all participants. In addition, to retain the anonymity of participants’ encrypted data from anyone including authorities, it exploits the re-encryption. Although the proposed framework also practices homomorphic encryption, which usually sacrifices the data integrity, here due to the usage of UTs, it is maintained. This paper is a preliminary report of the proposed framework.","PeriodicalId":166450,"journal":{"name":"2021 24th International Conference on Computer and Information Technology (ICCIT)","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114704218","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 : 2021-12-18DOI: 10.1109/ICCIT54785.2021.9689895
Azmain Yakin Srizon, Md. Ali Hossainy, Md Rakibul Haquez
Sign language is an essential tool for the deaf and the hard of hearing community of approximately 1.33 billion people. Due to this fact, researches have been conducted for decades for near-accurate recognition of sign characters. Sensor-based approaches and vision-based approaches have been adapted so far for tackling this dilemma. Sensor-based approaches can obtain high performance but it is costly and demands physical contact to sensors. On the other hand, vision-based approaches are not costly, need no contact but have not yet been able to produce a high accuracy like sensor-based approaches. The dilemma of sign characters recognition gets more problematic for Bengali sign language as not many datasets regarding Bengali sign language are available. Moreover, not many significant contributions can be found in this domain like other popular languages such as English, Turkish, Japanese, and Indian sign language. Furthermore, one of the most popular Bengali sign language datasets, Ishara-Lipi, consists of a few low-resolution samples. This study is focused on recognizing the low-resolution hand gestures of Bengali sign language. In this research, a convolutional neural network has been proposed which is suitable for the recognition of low-resolution sign gestures. Experimental results showed that the proposed approach achieved 99.08%, 99.38%, and 99.07% overall accuracy for digits, characters, and both digits and characters of the Ishara-Lipi dataset respectively.
{"title":"Low Resolution Hand Gestures Recognition of Bengali Sign Alphabet by Using a Convolutional Neural Network","authors":"Azmain Yakin Srizon, Md. Ali Hossainy, Md Rakibul Haquez","doi":"10.1109/ICCIT54785.2021.9689895","DOIUrl":"https://doi.org/10.1109/ICCIT54785.2021.9689895","url":null,"abstract":"Sign language is an essential tool for the deaf and the hard of hearing community of approximately 1.33 billion people. Due to this fact, researches have been conducted for decades for near-accurate recognition of sign characters. Sensor-based approaches and vision-based approaches have been adapted so far for tackling this dilemma. Sensor-based approaches can obtain high performance but it is costly and demands physical contact to sensors. On the other hand, vision-based approaches are not costly, need no contact but have not yet been able to produce a high accuracy like sensor-based approaches. The dilemma of sign characters recognition gets more problematic for Bengali sign language as not many datasets regarding Bengali sign language are available. Moreover, not many significant contributions can be found in this domain like other popular languages such as English, Turkish, Japanese, and Indian sign language. Furthermore, one of the most popular Bengali sign language datasets, Ishara-Lipi, consists of a few low-resolution samples. This study is focused on recognizing the low-resolution hand gestures of Bengali sign language. In this research, a convolutional neural network has been proposed which is suitable for the recognition of low-resolution sign gestures. Experimental results showed that the proposed approach achieved 99.08%, 99.38%, and 99.07% overall accuracy for digits, characters, and both digits and characters of the Ishara-Lipi dataset respectively.","PeriodicalId":166450,"journal":{"name":"2021 24th International Conference on Computer and Information Technology (ICCIT)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122892148","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}
An automated means for predicting the virus is of utmost importance to help the medical personnel to detect patients, prepare reports and produce results fast and impeccably so that people can get early treatment and prevent future transmissions. In this work, we proposed a COVID19 detection method using chest x-ray images by training and testing pre-trained deep neural network models, such as VGG19, InceptionV3, and Densenet201 individually, and got an accuracy of 96.9%, 95.2%, and 96.7% respectively. Then to bolster the performance of each model, we proposed an average weighted based ensemble approach and got an accuracy of 97.5%, which surpassed the performance of each separate model.
{"title":"An ensemble learning based approach to autonomous COVID19 detection using transfer learning with the help of pre-trained Deep Neural Network models","authors":"Faiza Anan Noor, Ishrakul Munzerin, A. Iqbal, Tanima Islam, Emam Hossain","doi":"10.1109/ICCIT54785.2021.9689825","DOIUrl":"https://doi.org/10.1109/ICCIT54785.2021.9689825","url":null,"abstract":"An automated means for predicting the virus is of utmost importance to help the medical personnel to detect patients, prepare reports and produce results fast and impeccably so that people can get early treatment and prevent future transmissions. In this work, we proposed a COVID19 detection method using chest x-ray images by training and testing pre-trained deep neural network models, such as VGG19, InceptionV3, and Densenet201 individually, and got an accuracy of 96.9%, 95.2%, and 96.7% respectively. Then to bolster the performance of each model, we proposed an average weighted based ensemble approach and got an accuracy of 97.5%, which surpassed the performance of each separate model.","PeriodicalId":166450,"journal":{"name":"2021 24th International Conference on Computer and Information Technology (ICCIT)","volume":"59 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126566539","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}
Question Answering demands a deep understanding of semantic relations among question, answer, and context. Multi-Task Learning (MTL) and Meta Learning with deep neural networks have recently shown impressive performance in many Natural Language Processing (NLP) tasks, particularly when there is inadequate data for training. But a little work has been done for a general NLP architecture that spans over many NLP tasks. In this paper, we present a model that can generalize to ten different NLP tasks. We demonstrate that multi-pointer-generator decoder and pre-trained language model is key to success and suppress all previous state-of-the-art baselines by 74 decaScore which is more than 12% absolute improvement over all of the datasets.
{"title":"Multitask Learning as Question Answering with BERT","authors":"Shishir Roy, Nayeem Ehtesham, Md Saiful Islam, Sabir Ismail","doi":"10.1109/ICCIT54785.2021.9689900","DOIUrl":"https://doi.org/10.1109/ICCIT54785.2021.9689900","url":null,"abstract":"Question Answering demands a deep understanding of semantic relations among question, answer, and context. Multi-Task Learning (MTL) and Meta Learning with deep neural networks have recently shown impressive performance in many Natural Language Processing (NLP) tasks, particularly when there is inadequate data for training. But a little work has been done for a general NLP architecture that spans over many NLP tasks. In this paper, we present a model that can generalize to ten different NLP tasks. We demonstrate that multi-pointer-generator decoder and pre-trained language model is key to success and suppress all previous state-of-the-art baselines by 74 decaScore which is more than 12% absolute improvement over all of the datasets.","PeriodicalId":166450,"journal":{"name":"2021 24th International Conference on Computer and Information Technology (ICCIT)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123130236","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 : 2021-12-18DOI: 10.1109/ICCIT54785.2021.9689813
Zahin Ahmed, Farishta Jayas Kinjol, I. Ananya
In recent years, the development of programming languages has been centered around making them easily understandable and learnable to users. Hence, the readability, writability of languages is being constantly improved while trying to keep the performance reliable. These factors affect how many new users start to use a particular language and how many experienced programmers continue to use it reliably in real applications. Hence, this research has compared the readability, writability, and reliability of six mainstream programming languages, namely C, C++, Java, JavaScript, Python, and R, based on their theoretical characteristics. Furthermore, we conducted a survey determining the choice of a language among programmers and nonprogrammers, which complemented the results gathered from the study. We found that Python outperforms others in terms of its readability and writability, while Java is proven to be the most reliable of all. We reported our findings, insights, and a discussion on the future development of better evaluation metrics.
{"title":"Comparative Analysis of Six Programming Languages Based on Readability, Writability, and Reliability","authors":"Zahin Ahmed, Farishta Jayas Kinjol, I. Ananya","doi":"10.1109/ICCIT54785.2021.9689813","DOIUrl":"https://doi.org/10.1109/ICCIT54785.2021.9689813","url":null,"abstract":"In recent years, the development of programming languages has been centered around making them easily understandable and learnable to users. Hence, the readability, writability of languages is being constantly improved while trying to keep the performance reliable. These factors affect how many new users start to use a particular language and how many experienced programmers continue to use it reliably in real applications. Hence, this research has compared the readability, writability, and reliability of six mainstream programming languages, namely C, C++, Java, JavaScript, Python, and R, based on their theoretical characteristics. Furthermore, we conducted a survey determining the choice of a language among programmers and nonprogrammers, which complemented the results gathered from the study. We found that Python outperforms others in terms of its readability and writability, while Java is proven to be the most reliable of all. We reported our findings, insights, and a discussion on the future development of better evaluation metrics.","PeriodicalId":166450,"journal":{"name":"2021 24th International Conference on Computer and Information Technology (ICCIT)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127675844","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 : 2021-12-18DOI: 10.1109/ICCIT54785.2021.9689883
M. N. Rahaman, M. S. Biswas, S. Chaki, M. M. Hossain, Shamim Ahmed, M. Biswas
Road region extraction is a crucial part of the vision-based driver assistance system of intelligent vehicles. This driver assistance system reduces road accidents, enhances safety, and improves traffic conditions. Autonomous Guided Vehicles are capable of performing required tasks in a defined environment without continuous human guidance. This research paper presents the design of a prototype autonomous guided vehicle which will detect and follow the lanes using the Probabilistic Hough Transform (PHT) algorithm. To do so, We convert our RGB road images into an HSV color model and then apply Gaussian smoothing to the converted grayscale image. For detection purposes, we process our region of interest (ROI) using a polygon clipping algorithm. Then, we apply Probabilistic Hough Transform upon the ROI image while setting all the parameters in our proposed lane detection algorithm. We present a robust real-time approach to extract road regions even in critical conditions like urban roads, unmarked roads. We have applied our proposed framework on the CALTECH dataset and gained 94.7% detection accuracy results in our experimental setup.
{"title":"Lane Detection for Autonomous Vehicle Management: PHT Approach","authors":"M. N. Rahaman, M. S. Biswas, S. Chaki, M. M. Hossain, Shamim Ahmed, M. Biswas","doi":"10.1109/ICCIT54785.2021.9689883","DOIUrl":"https://doi.org/10.1109/ICCIT54785.2021.9689883","url":null,"abstract":"Road region extraction is a crucial part of the vision-based driver assistance system of intelligent vehicles. This driver assistance system reduces road accidents, enhances safety, and improves traffic conditions. Autonomous Guided Vehicles are capable of performing required tasks in a defined environment without continuous human guidance. This research paper presents the design of a prototype autonomous guided vehicle which will detect and follow the lanes using the Probabilistic Hough Transform (PHT) algorithm. To do so, We convert our RGB road images into an HSV color model and then apply Gaussian smoothing to the converted grayscale image. For detection purposes, we process our region of interest (ROI) using a polygon clipping algorithm. Then, we apply Probabilistic Hough Transform upon the ROI image while setting all the parameters in our proposed lane detection algorithm. We present a robust real-time approach to extract road regions even in critical conditions like urban roads, unmarked roads. We have applied our proposed framework on the CALTECH dataset and gained 94.7% detection accuracy results in our experimental setup.","PeriodicalId":166450,"journal":{"name":"2021 24th International Conference on Computer and Information Technology (ICCIT)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121399608","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 : 2021-12-18DOI: 10.1109/ICCIT54785.2021.9689888
Md. Kaviul Hossain, Tasmim Promi, Piash Paul
Credit card fraudulence is a federal offense that takes place frequently in recent times. The phenomenon where an imposter or a scammer tries to make an illegal purchase or transfer of money from one account to another using a credit card that does not belong to him/her, is coined as Credit Card Fraudulence. In modern world, credit card fraud or any type of payment card fraud is a very common but serious crime that occurs both offline and online. But with the help of machine learning algorithms and Salient Feature Extraction Technique (SFET) we can easily detect such offense and help in further investigations. From time to time many data scientists, data analysts, machine learning engineers and other researchers have designed many algorithms to detect credit card frauds. By extracting the most relevant and important features of a transaction, it is quite possible to detect credit card fraud very quickly & efficiently. In this paper, we have shown such an improved way by using Adaptive Synthetic oversampling (ADASYN) model with five notable supervised machine learning models namely Random Forest, Support Vector Machine, Naive Bayes, Logistics Regression and K-Nearest Neighbour. Out of these five machine learning models, K-Nearest Neighbour has shown the best precision, recall, specificity & accuracy. The performance accuracy of Random Forest, Logistic Regression, K-Nearest Neighbour, Naive Bayes & Support Vector Machines are 96.04%, 81.31%, 96.22%, 79.22% & 50.06% respectively.
{"title":"Credit card fraudulence detection using Salient Feature Extraction Technique with Adaptive Synthetic Oversampling Models","authors":"Md. Kaviul Hossain, Tasmim Promi, Piash Paul","doi":"10.1109/ICCIT54785.2021.9689888","DOIUrl":"https://doi.org/10.1109/ICCIT54785.2021.9689888","url":null,"abstract":"Credit card fraudulence is a federal offense that takes place frequently in recent times. The phenomenon where an imposter or a scammer tries to make an illegal purchase or transfer of money from one account to another using a credit card that does not belong to him/her, is coined as Credit Card Fraudulence. In modern world, credit card fraud or any type of payment card fraud is a very common but serious crime that occurs both offline and online. But with the help of machine learning algorithms and Salient Feature Extraction Technique (SFET) we can easily detect such offense and help in further investigations. From time to time many data scientists, data analysts, machine learning engineers and other researchers have designed many algorithms to detect credit card frauds. By extracting the most relevant and important features of a transaction, it is quite possible to detect credit card fraud very quickly & efficiently. In this paper, we have shown such an improved way by using Adaptive Synthetic oversampling (ADASYN) model with five notable supervised machine learning models namely Random Forest, Support Vector Machine, Naive Bayes, Logistics Regression and K-Nearest Neighbour. Out of these five machine learning models, K-Nearest Neighbour has shown the best precision, recall, specificity & accuracy. The performance accuracy of Random Forest, Logistic Regression, K-Nearest Neighbour, Naive Bayes & Support Vector Machines are 96.04%, 81.31%, 96.22%, 79.22% & 50.06% respectively.","PeriodicalId":166450,"journal":{"name":"2021 24th International Conference on Computer and Information Technology (ICCIT)","volume":"117 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133265974","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 : 2021-12-18DOI: 10.1109/ICCIT54785.2021.9689866
Istiak Ahmed Mondal, Md. Enamul Haque, Al-Maruf Hassan, Swakkhar Shatabda
With the rising trend in online transactions, the threat of financial fraud is also rising. This makes the necessity for an effective Fraud Detection System (FDS) more than ever before. To develop such a system the financial institutes are moving towards machine learning-based approaches due to their effectiveness. Machine learning-based systems need historical data to learn. As fraud cases take place rarely, the number of positive labeled classes in financial fraud datasets are very small and the datasets remain imbalanced. For this, the possibility for machine learning-based FDS to produce misleading results is high. To counter this problem Machine Learning (ML) researchers use multiple solutions from the perspective of data-level approach, algorithm-level approach, feature engineering, ensemble models, or any combination of them. In this paper, we propose to use Generative Adversarial Network (GAN) based synthetic data generation to handle the data imbalance problem followed by an ensemble classifier for classification. We have used a standard benchmark dataset of credit card fraud data. In our experiments, we have used both traditional oversampling/undersampling and GAN-based techniques from the data-level approach and investigated their effectiveness using ML algorithms and ensemble models. We have found Generative Adversarial Network (GAN) to be more effective and stable in performance compared to traditional oversampling techniques for both ML and ensemble models. Experiments also suggest that the combination of GAN-based sampling and ensemble models provides the best results. We also have found Synthetic Minority Oversampling Technique (SMOTE) to provide more stable results compared to Adaptive Synthetic Sample (ADASYN) from the traditional oversampling technique.
{"title":"Handling Imbalanced Data for Credit Card Fraud Detection","authors":"Istiak Ahmed Mondal, Md. Enamul Haque, Al-Maruf Hassan, Swakkhar Shatabda","doi":"10.1109/ICCIT54785.2021.9689866","DOIUrl":"https://doi.org/10.1109/ICCIT54785.2021.9689866","url":null,"abstract":"With the rising trend in online transactions, the threat of financial fraud is also rising. This makes the necessity for an effective Fraud Detection System (FDS) more than ever before. To develop such a system the financial institutes are moving towards machine learning-based approaches due to their effectiveness. Machine learning-based systems need historical data to learn. As fraud cases take place rarely, the number of positive labeled classes in financial fraud datasets are very small and the datasets remain imbalanced. For this, the possibility for machine learning-based FDS to produce misleading results is high. To counter this problem Machine Learning (ML) researchers use multiple solutions from the perspective of data-level approach, algorithm-level approach, feature engineering, ensemble models, or any combination of them. In this paper, we propose to use Generative Adversarial Network (GAN) based synthetic data generation to handle the data imbalance problem followed by an ensemble classifier for classification. We have used a standard benchmark dataset of credit card fraud data. In our experiments, we have used both traditional oversampling/undersampling and GAN-based techniques from the data-level approach and investigated their effectiveness using ML algorithms and ensemble models. We have found Generative Adversarial Network (GAN) to be more effective and stable in performance compared to traditional oversampling techniques for both ML and ensemble models. Experiments also suggest that the combination of GAN-based sampling and ensemble models provides the best results. We also have found Synthetic Minority Oversampling Technique (SMOTE) to provide more stable results compared to Adaptive Synthetic Sample (ADASYN) from the traditional oversampling technique.","PeriodicalId":166450,"journal":{"name":"2021 24th International Conference on Computer and Information Technology (ICCIT)","volume":"83 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133071370","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 : 2021-12-18DOI: 10.1109/ICCIT54785.2021.9689778
Md Siam Talukder, M. Samsuzzaman, L. Paul, Md. Abdul Masud, R. Azim, M. Moniruzzaman
This paper explored a Broadband Corrugated Modified Vivaldi Antenna for head imaging applications using microwaves. The antenna is made up of a tapered feeding circular slotted patch and a rectangular modified wing shape ground with an electrical dimension of $0.49lambdatimes 0.41lambdatimes 0.009lambda$ at lower frequency of 1.67 GHz. Front and back radiator are placed on a substrate of FR-4 material having a permittivity of 4.3 and a thickness of 1.5 mm, which are less expensive and more widely accessible. The designed antenna works between 1.67 and 6.37 GHz, allowing the imaging system to function over a large fractional bandwidth of about 116.92 %. With an appropriate impedance matching and a directionally stable radiation characteristics, the antenna has a decent gain of 9 dBi. A 50$Omega$ microstrip line feeds the antenna. The simulation tools CST was used to evolve and optimize the design.
本文研究了一种用于微波头部成像的宽带波形改进维瓦尔第天线。该天线由锥形馈电圆形开槽贴片和矩形修正翼形地面组成，低频为1.67 GHz，电尺寸为$0.49lambdatimes 0.41lambdatimes 0.009lambda$。前后散热器放置在介电常数为4.3，厚度为1.5 mm的FR-4材料的基板上，这种材料更便宜，更容易获得。设计的天线工作在1.67和6.37 GHz之间，允许成像系统在大约116.92的大分数带宽上工作 %. With an appropriate impedance matching and a directionally stable radiation characteristics, the antenna has a decent gain of 9 dBi. A 50$Omega$ microstrip line feeds the antenna. The simulation tools CST was used to evolve and optimize the design.
{"title":"Broadband Corrugated Modified Vivaldi Antenna for Microwave based Imaging Applications","authors":"Md Siam Talukder, M. Samsuzzaman, L. Paul, Md. Abdul Masud, R. Azim, M. Moniruzzaman","doi":"10.1109/ICCIT54785.2021.9689778","DOIUrl":"https://doi.org/10.1109/ICCIT54785.2021.9689778","url":null,"abstract":"This paper explored a Broadband Corrugated Modified Vivaldi Antenna for head imaging applications using microwaves. The antenna is made up of a tapered feeding circular slotted patch and a rectangular modified wing shape ground with an electrical dimension of $0.49lambdatimes 0.41lambdatimes 0.009lambda$ at lower frequency of 1.67 GHz. Front and back radiator are placed on a substrate of FR-4 material having a permittivity of 4.3 and a thickness of 1.5 mm, which are less expensive and more widely accessible. The designed antenna works between 1.67 and 6.37 GHz, allowing the imaging system to function over a large fractional bandwidth of about 116.92 %. With an appropriate impedance matching and a directionally stable radiation characteristics, the antenna has a decent gain of 9 dBi. A 50$Omega$ microstrip line feeds the antenna. The simulation tools CST was used to evolve and optimize the design.","PeriodicalId":166450,"journal":{"name":"2021 24th International Conference on Computer and Information Technology (ICCIT)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133252016","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：