Pub Date : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10056009
Fatima Juairiah, Mostafa Mahatabe, Hasan Bin Jamal, Aysha Shiddika, Tanvir Rouf Shawon, Nibir Chandra Mandal
Predicting future stock volatility has always been a demanding chore for research studies. Individuals around the world have long regarded the stock market as a substantial profit. A stock data set contains numerous precise terms that are difficult for an individual to comprehend when considering stock market expenditures. An essential manifestation of a stock’s performance on the stock market is its closing price, but it is challenging to estimate the stock market’s price movements. This study aims to provide a future market scenario supported by statistical data. We used the Microsoft Corporation Stock dataset from 1986 to 2022. To foresee stock market volatility, we used time series analysis with the Long Short-Term Memory (LSTM), Bidirectional Long-short Term Memory (Bi-LSTM), Autoregressive Integrated Moving Average (ARIMA), Hidden Markov Model (HMM), and Multi-Head Attention. We have achieved 0.153, 0.202, 6.674, 14.760, and 21.493 for Transformer, HMM, ARIMA, BiLSTM, and LSTM respectively.
{"title":"Stock Price Prediction: A Time Series Analysis","authors":"Fatima Juairiah, Mostafa Mahatabe, Hasan Bin Jamal, Aysha Shiddika, Tanvir Rouf Shawon, Nibir Chandra Mandal","doi":"10.1109/ICCIT57492.2022.10056009","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10056009","url":null,"abstract":"Predicting future stock volatility has always been a demanding chore for research studies. Individuals around the world have long regarded the stock market as a substantial profit. A stock data set contains numerous precise terms that are difficult for an individual to comprehend when considering stock market expenditures. An essential manifestation of a stock’s performance on the stock market is its closing price, but it is challenging to estimate the stock market’s price movements. This study aims to provide a future market scenario supported by statistical data. We used the Microsoft Corporation Stock dataset from 1986 to 2022. To foresee stock market volatility, we used time series analysis with the Long Short-Term Memory (LSTM), Bidirectional Long-short Term Memory (Bi-LSTM), Autoregressive Integrated Moving Average (ARIMA), Hidden Markov Model (HMM), and Multi-Head Attention. We have achieved 0.153, 0.202, 6.674, 14.760, and 21.493 for Transformer, HMM, ARIMA, BiLSTM, and LSTM respectively.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127954092","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 : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10055792
M. Rana, Shaikh Anowarul
Brain-computer interface (BCI) systems have recently gained much attention due to their ability to interpret human thought processes from electroencephalogram (EEG)-based motor imagery (MI) signals. Among various machine learning based methods, instead of using hand-crafted features, end-to-end deep learning (DL) based methods are getting popularity due to their very satisfactory performance. Even in case of non-stationary and subject specific MI-EEG signals, DL methods offer high accuracy in classifying MI tasks. In this paper, a dual band convolutional neural network (DBCNN) based on MI-EEG two band limited efficient signals is proposed. Here, beta wave signals are used to improve classification performance in motor imagery tasks. Beta waves are associated with motor imagery tasks. On the other hand our second signal is the pre-processed bandpass signal. The DBCNN model uses temporal CNN and spatial CNN to explore temporal and spatial features. This model also uses a spatial convolutional network for single channels from multiple channels. Finally the data are classified for MI-EEG signal classification accuracy. An extensive experiment on MI-based BCI IV 2a EEG dataset on nine subjects shows that very satisfactory classification performance is achieved. The proposed model offers higher classification accuracy that is obtained by some state-of-the-art methods.
脑机接口(BCI)系统由于能够从基于脑电图(EEG)的运动图像(MI)信号中解释人类的思维过程，最近受到了广泛的关注。在各种基于机器学习的方法中，基于端到端深度学习(DL)的方法由于其非常令人满意的性能而越来越受欢迎，而不是使用手工制作的特征。即使在非平稳和受试者特定的MI- eeg信号情况下，深度学习方法对MI任务的分类也具有很高的准确性。本文提出了一种基于MI-EEG两波段有限有效信号的双波段卷积神经网络(DBCNN)。在这里，β波信号被用来提高运动想象任务的分类性能。β波与运动想象任务有关。另一方面，我们的第二个信号是预处理带通信号。DBCNN模型使用时间CNN和空间CNN来探索时空特征。该模型还使用空间卷积网络对多个通道中的单个通道进行处理。最后对数据进行分类，提高脑电信号的分类精度。在基于mi的BCI IV 2a脑电数据集上对9个被试进行了大量实验，结果表明该方法取得了令人满意的分类效果。所提出的模型具有比现有方法更高的分类精度。
{"title":"An efficient Dual-Band CNN for Motor Imagery EEG Signal Classification","authors":"M. Rana, Shaikh Anowarul","doi":"10.1109/ICCIT57492.2022.10055792","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055792","url":null,"abstract":"Brain-computer interface (BCI) systems have recently gained much attention due to their ability to interpret human thought processes from electroencephalogram (EEG)-based motor imagery (MI) signals. Among various machine learning based methods, instead of using hand-crafted features, end-to-end deep learning (DL) based methods are getting popularity due to their very satisfactory performance. Even in case of non-stationary and subject specific MI-EEG signals, DL methods offer high accuracy in classifying MI tasks. In this paper, a dual band convolutional neural network (DBCNN) based on MI-EEG two band limited efficient signals is proposed. Here, beta wave signals are used to improve classification performance in motor imagery tasks. Beta waves are associated with motor imagery tasks. On the other hand our second signal is the pre-processed bandpass signal. The DBCNN model uses temporal CNN and spatial CNN to explore temporal and spatial features. This model also uses a spatial convolutional network for single channels from multiple channels. Finally the data are classified for MI-EEG signal classification accuracy. An extensive experiment on MI-based BCI IV 2a EEG dataset on nine subjects shows that very satisfactory classification performance is achieved. The proposed model offers higher classification accuracy that is obtained by some state-of-the-art methods.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129885918","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 : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10055190
Suvom Shaha, F. Shah, Amir Hossain Raj, Ashek Seum, Saiful Islam, Sifat Ahmed
Generating automatic video captions is one of the most challenging Artificial Intelligence tasks as it combines Computer Vision and Natural Language Processing research areas. The task is more difficult for a complex language like Bengali as there is a general lack of video captioning datasets in the Bengali language. To overcome this challenge, we introduce a fully human-annotated dataset of Bengali captions in this research for the videos of the MSVD dataset. We have proposed a novel end-to-end architecture with an attention-based decoder to generate meaningful video captions in the Bengali language. First, spatial and temporal features of videos are combined using Bidirectional Gated Recurrent Units (Bi-GRU) that generate the input feature, which is later fed to the attention layer along with embedded caption features. This attention mechanism explores the interdependence between visual and textual representations. Then, a double-layered GRU takes these combined attention features for generating meaningful sentences. We trained this model on our proposed dataset and achieved 39.35% in BLEU-4, 59.67% in CIDEr, and 65.34% score in ROUGE. This is the state-of-the-art result compared to any other video captioning work available in the Bengali language.
{"title":"Video Captioning in Bengali With Visual Attention","authors":"Suvom Shaha, F. Shah, Amir Hossain Raj, Ashek Seum, Saiful Islam, Sifat Ahmed","doi":"10.1109/ICCIT57492.2022.10055190","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055190","url":null,"abstract":"Generating automatic video captions is one of the most challenging Artificial Intelligence tasks as it combines Computer Vision and Natural Language Processing research areas. The task is more difficult for a complex language like Bengali as there is a general lack of video captioning datasets in the Bengali language. To overcome this challenge, we introduce a fully human-annotated dataset of Bengali captions in this research for the videos of the MSVD dataset. We have proposed a novel end-to-end architecture with an attention-based decoder to generate meaningful video captions in the Bengali language. First, spatial and temporal features of videos are combined using Bidirectional Gated Recurrent Units (Bi-GRU) that generate the input feature, which is later fed to the attention layer along with embedded caption features. This attention mechanism explores the interdependence between visual and textual representations. Then, a double-layered GRU takes these combined attention features for generating meaningful sentences. We trained this model on our proposed dataset and achieved 39.35% in BLEU-4, 59.67% in CIDEr, and 65.34% score in ROUGE. This is the state-of-the-art result compared to any other video captioning work available in the Bengali language.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"55 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115188309","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 : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10055237
Sabiha Islam, Md. Shafiul Alam Forhad, Hasan Murad
Nowadays, people are constantly expressing their emotions, thoughts, opinions, and daily activities on social media. Therefore, social media posts have become a powerful tool among psychiatrists for the early detection of suicidal tendencies. However, the automatic detection of suicidal posts has become a challenging problem among researchers in the field of Natural Language Processing (NLP). A significant number of previous works have been found in the literature for the automatic detection of suicidal posts in different languages such as English. However, little effort has been devoted to automatically detecting suicidal posts in low-resource languages like Bangla due to the lack of available datasets. In this study, we have created a noble Bangla suicidal posts dataset named BanglaSPD and compared the performance of various machine learning and deep learning models for suicide attempt prediction by training and evaluating with the dataset. Finally, we have found that a deep learning-based model with CNN+BiLSTM has outperformed with 0.61 F1-Score in Fasttext word embedding methods.
{"title":"BanglaSAPM: A Deep Learning Model for Suicidal Attempt Prediction Using Social Media Content in Bangla","authors":"Sabiha Islam, Md. Shafiul Alam Forhad, Hasan Murad","doi":"10.1109/ICCIT57492.2022.10055237","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055237","url":null,"abstract":"Nowadays, people are constantly expressing their emotions, thoughts, opinions, and daily activities on social media. Therefore, social media posts have become a powerful tool among psychiatrists for the early detection of suicidal tendencies. However, the automatic detection of suicidal posts has become a challenging problem among researchers in the field of Natural Language Processing (NLP). A significant number of previous works have been found in the literature for the automatic detection of suicidal posts in different languages such as English. However, little effort has been devoted to automatically detecting suicidal posts in low-resource languages like Bangla due to the lack of available datasets. In this study, we have created a noble Bangla suicidal posts dataset named BanglaSPD and compared the performance of various machine learning and deep learning models for suicide attempt prediction by training and evaluating with the dataset. Finally, we have found that a deep learning-based model with CNN+BiLSTM has outperformed with 0.61 F1-Score in Fasttext word embedding methods.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115288390","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}
Controlling traffic congestion is one of the most challenging issues in the real world, and an automated traffic monitoring system established on the computer vision and internet of things is an emerging scientific field in research. Traffic Congestion becomes severe in the most densely populated area to move timely from one place to another, passing through massive traffic. Although several research papers have presented their solutions for monitoring traffic congestion, we do not receive a permanent solution. In this paper, We try to find a different but unique way to overcome it. Our proposed system works with real-time video sequential data and detects different vehicle shapes using contour-based learning and a convex hull algorithm. We execute the procedure in a way where we work with per-second frames (fps) of the video, and then we determine the total, running, and motionless stagnant vehicles in each frame. After observing vehicle differentiation, an automated decision will be generated, i.e., JAM or NOT JAM which will be transmitted to the Map for users’ assistance. After that, the decision about the traffic situation of all nodes makes a combination, and then we find two patterns, one will identify the traffic situation of all connected roads, and another will control the traffic signal automatically.
{"title":"Road Traffic Update and Control Traffic Light using Dynamic Patterns from Video Streaming Data","authors":"Md. Shafkat Islam, Md. Maharub Hossain, Mizbatul Jannat Refat","doi":"10.1109/ICCIT57492.2022.10055520","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055520","url":null,"abstract":"Controlling traffic congestion is one of the most challenging issues in the real world, and an automated traffic monitoring system established on the computer vision and internet of things is an emerging scientific field in research. Traffic Congestion becomes severe in the most densely populated area to move timely from one place to another, passing through massive traffic. Although several research papers have presented their solutions for monitoring traffic congestion, we do not receive a permanent solution. In this paper, We try to find a different but unique way to overcome it. Our proposed system works with real-time video sequential data and detects different vehicle shapes using contour-based learning and a convex hull algorithm. We execute the procedure in a way where we work with per-second frames (fps) of the video, and then we determine the total, running, and motionless stagnant vehicles in each frame. After observing vehicle differentiation, an automated decision will be generated, i.e., JAM or NOT JAM which will be transmitted to the Map for users’ assistance. After that, the decision about the traffic situation of all nodes makes a combination, and then we find two patterns, one will identify the traffic situation of all connected roads, and another will control the traffic signal automatically.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132072250","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 : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10055315
Md. Motaleb Hossen Manik, Kazi Md. Rokibul Alam, Y. Morimoto
For commercial purposes, business organizations with mutual interests while competing with one another usually intend to keep their datasets to be confidential. Secure multiparty skyline queries enable these organizations to retrieve dominated data objects from their sensitive datasets. Here, a distinction among datasets according to priority, retaining data privacy along with integrity, targeted enquires, etc., are crucial. This paper proposes a secure framework for multi-party skyline queries that incorporates these requirements together. To prioritize datasets it assigns distinct weights to parties’ datasets. To retain adequate privacy it exploits a cryptosystem that engages a single key for encryption but multiple separate keys for decryption. To attain data anonymity it re-encrypts and shuffles the encrypted data. To assure data integrity it deploys a unique block comprising encrypted data designated for every party within the blockchain storage. To enable enlisted users to ask intended queries it accepts their inquiries. This paper is a preliminary report and evaluation of the proposed framework.
{"title":"A Blockchain Based Secure Framework for User-centric Multi-party Skyline Queries","authors":"Md. Motaleb Hossen Manik, Kazi Md. Rokibul Alam, Y. Morimoto","doi":"10.1109/ICCIT57492.2022.10055315","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055315","url":null,"abstract":"For commercial purposes, business organizations with mutual interests while competing with one another usually intend to keep their datasets to be confidential. Secure multiparty skyline queries enable these organizations to retrieve dominated data objects from their sensitive datasets. Here, a distinction among datasets according to priority, retaining data privacy along with integrity, targeted enquires, etc., are crucial. This paper proposes a secure framework for multi-party skyline queries that incorporates these requirements together. To prioritize datasets it assigns distinct weights to parties’ datasets. To retain adequate privacy it exploits a cryptosystem that engages a single key for encryption but multiple separate keys for decryption. To attain data anonymity it re-encrypts and shuffles the encrypted data. To assure data integrity it deploys a unique block comprising encrypted data designated for every party within the blockchain storage. To enable enlisted users to ask intended queries it accepts their inquiries. This paper is a preliminary report and evaluation of the proposed framework.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"76 7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130620013","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 : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10054823
Anonnya Ghosh, Raqeebir Rab, Ashikur Rahman
Since 1906 till today animation has been a popular method of storytelling. Animated characters portray unique stories in a more perceivable way by expressing diverse facial expressions. Detection of these emotions has not yet gained as much acclaim as detection of human facial expressions. This research aims to classify and predict emotions of animated characters using Deep Learning Techniques. Images of seven facial expressions namely Anger, Disgust, Fear, Joy, Neutral, Sadness and Surprise are classified using Residual Network (ResNet) and Transfer Learning. In order to conduct research on emotion identification in animated faces, we generated a new dataset with fewer images than the existing dataset [1]. Features from the faces are extracted using Uniform Local Binary Patterns (LBP) and fed to Convolutional Neural Network (CNN) model for classification. In our proposed models, the Transfer Learning-based ResNet50 and ResNet101,were trained on the ImageNet [10] dataset. Among the models ResNet101 achieved the highest detection accuracy of 94% and ResNet50 showed lowest time complexity.
{"title":"Transfer learning based Facial Emotion Detection for Animated Characters","authors":"Anonnya Ghosh, Raqeebir Rab, Ashikur Rahman","doi":"10.1109/ICCIT57492.2022.10054823","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10054823","url":null,"abstract":"Since 1906 till today animation has been a popular method of storytelling. Animated characters portray unique stories in a more perceivable way by expressing diverse facial expressions. Detection of these emotions has not yet gained as much acclaim as detection of human facial expressions. This research aims to classify and predict emotions of animated characters using Deep Learning Techniques. Images of seven facial expressions namely Anger, Disgust, Fear, Joy, Neutral, Sadness and Surprise are classified using Residual Network (ResNet) and Transfer Learning. In order to conduct research on emotion identification in animated faces, we generated a new dataset with fewer images than the existing dataset [1]. Features from the faces are extracted using Uniform Local Binary Patterns (LBP) and fed to Convolutional Neural Network (CNN) model for classification. In our proposed models, the Transfer Learning-based ResNet50 and ResNet101,were trained on the ImageNet [10] dataset. Among the models ResNet101 achieved the highest detection accuracy of 94% and ResNet50 showed lowest time complexity.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"512 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123570202","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 : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10055716
Syed Taha Yeasin Ramadan, T. Sakib, Md. Mizba Ul Haque, N. Sharmin, Md. Mahbubur Rahman
Rice is one of the most produced crops in the world and the staple food in many South Asian countries. Rice leaf disease can affect the production of rice vastly, which can be prevented through the early detection of it. Many machine learning techniques have been used in recent years to help in the prevention of one of the most serious concerns, which is disease transmission.But there are limited images available of diseased leaf compared to healthy images which makes life tougher for machine learning models as they need a good amount of data for training. To solve this problem, a Generative adversarial network(GAN) has been used in recent days to create new, synthetic instances of an image that can pass as a real image. Recently, it has been used widely in the field of leaf disease identification. But there is very limited work done on rice diseases. In this paper, SRGAN (Super Resolution-GAN) has been considered as a data augmentation method to balance the dataset. Afterward, DenseNet121, DenseNet169, MobileNetV2, and VGG16 have been applied to classify the diseases. Experiment results show that the newly created augmented dataset produces the best results with both DenseNet169 and moboleNetV2 when compared to other models, with high accuracy of 94.30.
{"title":"Generative Adversarial Network-based Augmented Rice Leaf Disease Detection using Deep Learning","authors":"Syed Taha Yeasin Ramadan, T. Sakib, Md. Mizba Ul Haque, N. Sharmin, Md. Mahbubur Rahman","doi":"10.1109/ICCIT57492.2022.10055716","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055716","url":null,"abstract":"Rice is one of the most produced crops in the world and the staple food in many South Asian countries. Rice leaf disease can affect the production of rice vastly, which can be prevented through the early detection of it. Many machine learning techniques have been used in recent years to help in the prevention of one of the most serious concerns, which is disease transmission.But there are limited images available of diseased leaf compared to healthy images which makes life tougher for machine learning models as they need a good amount of data for training. To solve this problem, a Generative adversarial network(GAN) has been used in recent days to create new, synthetic instances of an image that can pass as a real image. Recently, it has been used widely in the field of leaf disease identification. But there is very limited work done on rice diseases. In this paper, SRGAN (Super Resolution-GAN) has been considered as a data augmentation method to balance the dataset. Afterward, DenseNet121, DenseNet169, MobileNetV2, and VGG16 have been applied to classify the diseases. Experiment results show that the newly created augmented dataset produces the best results with both DenseNet169 and moboleNetV2 when compared to other models, with high accuracy of 94.30.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"131 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126728004","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 : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10055395
Md. Moynul Asik Moni, Maharshi Niloy, Aquibul Haq Chowdhury, Farah Jasmin Khan, Md. Fahmid-Ul-Alam Juboraj, Amitabha Chakrabarty
Process scheduling is an integral part of operating systems. The most widely used scheduling algorithm in operating systems is Round Robin, but the average waiting time in RR is often quite long. The purpose of this study is to propose a new algorithm to minimize waiting time and process starvation by determining the optimal time quantum by predicting CPU burst time. For burst time prediction, we are using the machine learning algorithms like linear regression, decision tree, k-nearest neighbors, and Neural Network Model Multi-Layer Perceptron. Moreover, for 10000 predicted burst time of processes with the same configuration, we have compared the average turnaround time, the average waiting time and the number of context switches of the proposed modified round robin algorithm with Traditional Round Robin, Modified Round Robin, Optimized Round Robin and Self-Adjustment Round Robin. The proposed modified round robin i.e. Absolute Difference Based Time Quantum Round Robin (ADRR) is found to be almost 2 times faster than the other algorithm in terms of process scheduling for the used dataset which contains a huge load of processes.
{"title":"Comparative Analysis of Process Scheduling Algorithm using AI models","authors":"Md. Moynul Asik Moni, Maharshi Niloy, Aquibul Haq Chowdhury, Farah Jasmin Khan, Md. Fahmid-Ul-Alam Juboraj, Amitabha Chakrabarty","doi":"10.1109/ICCIT57492.2022.10055395","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055395","url":null,"abstract":"Process scheduling is an integral part of operating systems. The most widely used scheduling algorithm in operating systems is Round Robin, but the average waiting time in RR is often quite long. The purpose of this study is to propose a new algorithm to minimize waiting time and process starvation by determining the optimal time quantum by predicting CPU burst time. For burst time prediction, we are using the machine learning algorithms like linear regression, decision tree, k-nearest neighbors, and Neural Network Model Multi-Layer Perceptron. Moreover, for 10000 predicted burst time of processes with the same configuration, we have compared the average turnaround time, the average waiting time and the number of context switches of the proposed modified round robin algorithm with Traditional Round Robin, Modified Round Robin, Optimized Round Robin and Self-Adjustment Round Robin. The proposed modified round robin i.e. Absolute Difference Based Time Quantum Round Robin (ADRR) is found to be almost 2 times faster than the other algorithm in terms of process scheduling for the used dataset which contains a huge load of processes.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"86 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114505270","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 : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10055205
Mahamudul Hasan Rafi, Shifat Islam, S. M. Hasan Imtiaz Labib, SM Sajid Hasan, F. Shah, Sifat Ahmed
Visual Question Answering (VQA) is a challenging task in Artificial Intelligence (AI), where an AI agent answers questions regarding visual content based on images provided. Therefore, to implement a VQA system, a computer system requires complex reasoning over visual aspects of images and textual parts of the questions to anticipate the correct answer. Although there is a good deal of VQA research in English, Bengali still needs to thoroughly explore this area of artificial intelligence. To address this, we have constructed a Bengali VQA dataset by preparing human-annotated question-answers using a small portion of the images from the VQA v2.0 dataset. To overcome high linguistic priors that hide the importance of precise visual information in visual question answering, we have used real-life scenarios to construct a balanced Bengali VQA dataset. This is the first human-annotated dataset of this kind in Bengali. We have proposed a Top-Down Attention-based approach in this study and conducted several studies to assess our model’s performance.
{"title":"A Deep Learning-Based Bengali Visual Question Answering System","authors":"Mahamudul Hasan Rafi, Shifat Islam, S. M. Hasan Imtiaz Labib, SM Sajid Hasan, F. Shah, Sifat Ahmed","doi":"10.1109/ICCIT57492.2022.10055205","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055205","url":null,"abstract":"Visual Question Answering (VQA) is a challenging task in Artificial Intelligence (AI), where an AI agent answers questions regarding visual content based on images provided. Therefore, to implement a VQA system, a computer system requires complex reasoning over visual aspects of images and textual parts of the questions to anticipate the correct answer. Although there is a good deal of VQA research in English, Bengali still needs to thoroughly explore this area of artificial intelligence. To address this, we have constructed a Bengali VQA dataset by preparing human-annotated question-answers using a small portion of the images from the VQA v2.0 dataset. To overcome high linguistic priors that hide the importance of precise visual information in visual question answering, we have used real-life scenarios to construct a balanced Bengali VQA dataset. This is the first human-annotated dataset of this kind in Bengali. We have proposed a Top-Down Attention-based approach in this study and conducted several studies to assess our model’s performance.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115918267","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：