Abstract Deep learning (DL) has revolutionized advanced digital picture processing, enabling significant advancements in computer vision (CV). However, it is important to note that older CV techniques, developed prior to the emergence of DL, still hold value and relevance. Particularly in the realm of more complex, three-dimensional (3D) data such as video and 3D models, CV and multimedia retrieval remain at the forefront of technological advancements. We provide critical insights into the progress made in developing higher-dimensional qualities through the application of DL, and also discuss the advantages and strategies employed in DL. With the widespread use of 3D sensor data and 3D modeling, the analysis and representation of the world in three dimensions have become commonplace. This progress has been facilitated by the development of additional sensors, driven by advancements in areas such as 3D gaming and self-driving vehicles. These advancements have enabled researchers to create feature description models that surpass traditional two-dimensional approaches. This study reveals the current state of advanced digital picture processing, highlighting the role of DL in pushing the boundaries of CV and multimedia retrieval in handling complex, 3D data.
{"title":"Development and research of deep neural network fusion computer vision technology","authors":"Jiangtao Wang","doi":"10.1515/jisys-2022-0264","DOIUrl":"https://doi.org/10.1515/jisys-2022-0264","url":null,"abstract":"Abstract Deep learning (DL) has revolutionized advanced digital picture processing, enabling significant advancements in computer vision (CV). However, it is important to note that older CV techniques, developed prior to the emergence of DL, still hold value and relevance. Particularly in the realm of more complex, three-dimensional (3D) data such as video and 3D models, CV and multimedia retrieval remain at the forefront of technological advancements. We provide critical insights into the progress made in developing higher-dimensional qualities through the application of DL, and also discuss the advantages and strategies employed in DL. With the widespread use of 3D sensor data and 3D modeling, the analysis and representation of the world in three dimensions have become commonplace. This progress has been facilitated by the development of additional sensors, driven by advancements in areas such as 3D gaming and self-driving vehicles. These advancements have enabled researchers to create feature description models that surpass traditional two-dimensional approaches. This study reveals the current state of advanced digital picture processing, highlighting the role of DL in pushing the boundaries of CV and multimedia retrieval in handling complex, 3D data.","PeriodicalId":46139,"journal":{"name":"Journal of Intelligent Systems","volume":"148 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135157947","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}
Abstract In the era of big data, data information has exploded, and all walks of life are impacted by big data. The arrival of big data provides the possibility for the realization of intelligent financial analysis of enterprises. At present, most enterprises’ financial analysis and decision-making based on the analysis results are mainly based on human resources, with poor automation and obvious problems in efficiency and error. In order to help the senior management of enterprises to conduct scientific and effective management, the study uses big data web crawler technology and ETL technology to process data and build an intelligent financial decision support system integrating big data together with Internet plus platform. J Group in S Province is taken as an example to study the effect before and after the application of intelligent financial decision support system. The results show that the crawler technology can monitor the basic data and the big data in the industry in real time, and improve the accuracy of decision-making. Through the intelligent financial decision support system which integrates big data, the core indexes such as profit, net asset return, and accounts receivable of the enterprises can be clearly displayed. The system can query the causes of financial changes hidden behind the financial data. Through the intelligent financial decision support system, it is found that the asset liability ratio, current assets growth rate, operating income growth rate, and financial expenses of J Group are 55.27, 10.38, 20.28%, and 1,974 million RMB, respectively. The growth rate of real sales income of J Group is 0.63%, which is 31.27% less than the excellent value of the industry 31.90%. After adopting the intelligent financial decision support system, the monthly financial statements of the enterprises increase significantly, and the monthly report analysis time decreases. The maximum number of financial statements received by the Group per month is 332, and the processing time at this time is only 2 h. According to the results, it can be seen that the intelligent financial decision support system integrating big data as the research result can effectively improve the financial management level of enterprises, improve the usefulness of financial decision-making, and make practical contributions to the field of corporate financial decision-making.
{"title":"Intelligent financial decision support system based on big data","authors":"Danna Tong, Guixian Tian","doi":"10.1515/jisys-2022-0320","DOIUrl":"https://doi.org/10.1515/jisys-2022-0320","url":null,"abstract":"Abstract In the era of big data, data information has exploded, and all walks of life are impacted by big data. The arrival of big data provides the possibility for the realization of intelligent financial analysis of enterprises. At present, most enterprises’ financial analysis and decision-making based on the analysis results are mainly based on human resources, with poor automation and obvious problems in efficiency and error. In order to help the senior management of enterprises to conduct scientific and effective management, the study uses big data web crawler technology and ETL technology to process data and build an intelligent financial decision support system integrating big data together with Internet plus platform. J Group in S Province is taken as an example to study the effect before and after the application of intelligent financial decision support system. The results show that the crawler technology can monitor the basic data and the big data in the industry in real time, and improve the accuracy of decision-making. Through the intelligent financial decision support system which integrates big data, the core indexes such as profit, net asset return, and accounts receivable of the enterprises can be clearly displayed. The system can query the causes of financial changes hidden behind the financial data. Through the intelligent financial decision support system, it is found that the asset liability ratio, current assets growth rate, operating income growth rate, and financial expenses of J Group are 55.27, 10.38, 20.28%, and 1,974 million RMB, respectively. The growth rate of real sales income of J Group is 0.63%, which is 31.27% less than the excellent value of the industry 31.90%. After adopting the intelligent financial decision support system, the monthly financial statements of the enterprises increase significantly, and the monthly report analysis time decreases. The maximum number of financial statements received by the Group per month is 332, and the processing time at this time is only 2 h. According to the results, it can be seen that the intelligent financial decision support system integrating big data as the research result can effectively improve the financial management level of enterprises, improve the usefulness of financial decision-making, and make practical contributions to the field of corporate financial decision-making.","PeriodicalId":46139,"journal":{"name":"Journal of Intelligent Systems","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135358255","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}
Abstract Paternity testing using a deoxyribose nucleic acid (DNA) profile is an essential branch of forensic science, and DNA short tandem repeat (STR) is usually used for this purpose. Nowadays, in third-world countries, conventional kinship analysis techniques used in forensic investigations result in inadequate accuracy measurements, especially when dealing with large human STR datasets; they compare human profiles manually so that the number of samples is limited due to the required human efforts and time consumption. By utilizing automation made possible by AI, forensic investigations are conducted more efficiently, saving both time conception and cost. In this article, we propose a new algorithm for predicting paternity based on the 15-loci STR-DNA datasets using a deep neural network (DNN), where comparisons among many human profiles are held regardless of the limitation of the number of samples. For the purpose of paternity testing, familial data are artificially created based on the real data of individual Iraqi people from Al-Najaf province. Such action helps to overcome the shortage of Iraqi data due to restricted policies and the secrecy of familial datasets. About 53,530 datasets are used in the proposed DNN model for the purpose of training and testing. The Keras library based on Python is used to implement and test the proposed system, as well as the confusion matrix and receiver operating characteristic curve for system evaluation. The system shows excellent accuracy of 99.6% in paternity tests, which is the highest accuracy compared to the existing works. This system shows a good attempt at testing paternity based on a technique of artificial intelligence.
{"title":"A deep neural network model for paternity testing based on 15-loci STR for Iraqi families","authors":"Donya A. Khalid, Nasser Nafea","doi":"10.1515/jisys-2023-0041","DOIUrl":"https://doi.org/10.1515/jisys-2023-0041","url":null,"abstract":"Abstract Paternity testing using a deoxyribose nucleic acid (DNA) profile is an essential branch of forensic science, and DNA short tandem repeat (STR) is usually used for this purpose. Nowadays, in third-world countries, conventional kinship analysis techniques used in forensic investigations result in inadequate accuracy measurements, especially when dealing with large human STR datasets; they compare human profiles manually so that the number of samples is limited due to the required human efforts and time consumption. By utilizing automation made possible by AI, forensic investigations are conducted more efficiently, saving both time conception and cost. In this article, we propose a new algorithm for predicting paternity based on the 15-loci STR-DNA datasets using a deep neural network (DNN), where comparisons among many human profiles are held regardless of the limitation of the number of samples. For the purpose of paternity testing, familial data are artificially created based on the real data of individual Iraqi people from Al-Najaf province. Such action helps to overcome the shortage of Iraqi data due to restricted policies and the secrecy of familial datasets. About 53,530 datasets are used in the proposed DNN model for the purpose of training and testing. The Keras library based on Python is used to implement and test the proposed system, as well as the confusion matrix and receiver operating characteristic curve for system evaluation. The system shows excellent accuracy of 99.6% in paternity tests, which is the highest accuracy compared to the existing works. This system shows a good attempt at testing paternity based on a technique of artificial intelligence.","PeriodicalId":46139,"journal":{"name":"Journal of Intelligent Systems","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135611752","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}
Abstract The nonlinear system is difficult to achieve the desired effect by using traditional proportional integral derivative (PID) or linear controller. First, this study presents an improved lazy learning algorithm based on k-vector nearest neighbors, which not only considers the matching of input and output data, but also considers the consistency of the model. Based on the optimization index of an additional penalty function, the optimal solution of the lazy learning is obtained by the iterative least-square method. Second, based on the improved lazy learning, an adaptive PID control algorithm is proposed. Finally, the control effect under the condition of complete data and incomplete data is compared by simulation experiment.
{"title":"Design model-free adaptive PID controller based on lazy learning algorithm","authors":"Hongcheng Zhou","doi":"10.1515/jisys-2022-0279","DOIUrl":"https://doi.org/10.1515/jisys-2022-0279","url":null,"abstract":"Abstract The nonlinear system is difficult to achieve the desired effect by using traditional proportional integral derivative (PID) or linear controller. First, this study presents an improved lazy learning algorithm based on k-vector nearest neighbors, which not only considers the matching of input and output data, but also considers the consistency of the model. Based on the optimization index of an additional penalty function, the optimal solution of the lazy learning is obtained by the iterative least-square method. Second, based on the improved lazy learning, an adaptive PID control algorithm is proposed. Finally, the control effect under the condition of complete data and incomplete data is compared by simulation experiment.","PeriodicalId":46139,"journal":{"name":"Journal of Intelligent Systems","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135953356","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}
Abstract In recent years, there have been several calls by practitioners of machine learning to provide more guidelines on how to use its methods and techniques. For example, the current literature on resampling methods is confusing and sometimes contradictory; worse, there are sometimes no practical guidelines offered at all. To address this shortcoming, a simulation study was conducted that evaluated ridge regression models fitted on five real-world datasets. The study compared the performance of four resampling methods, namely, Monte Carlo resampling, bootstrap, k-fold cross-validation, and repeated k-fold cross-validation. The goal was to find the best-fitting λ (regularization) parameter that would minimize mean squared error, by using nine variations of these resampling methods. For each of the nine resampling variations, 1,000 runs were performed to see how often a good fit, average fit, and poor fit λ value would be chosen. The resampling method that chose good fit values the greatest number of times was deemed the best method. Based on the results of the investigation, three general recommendations are made: (1) repeated k-fold cross-validation is the best method to select as a general-purpose resampling method; (2) k = 10 folds is a good choice in k-fold cross-validation; (3) Monte Carlo and bootstrap are underperformers, so they are not recommended as general-purpose resampling methods. At the same time, no resampling method was found to be uniformly better than the others.
{"title":"Validation of machine learning ridge regression models using Monte Carlo, bootstrap, and variations in cross-validation","authors":"Robbie T. Nakatsu","doi":"10.1515/jisys-2022-0224","DOIUrl":"https://doi.org/10.1515/jisys-2022-0224","url":null,"abstract":"Abstract In recent years, there have been several calls by practitioners of machine learning to provide more guidelines on how to use its methods and techniques. For example, the current literature on resampling methods is confusing and sometimes contradictory; worse, there are sometimes no practical guidelines offered at all. To address this shortcoming, a simulation study was conducted that evaluated ridge regression models fitted on five real-world datasets. The study compared the performance of four resampling methods, namely, Monte Carlo resampling, bootstrap, k-fold cross-validation, and repeated k-fold cross-validation. The goal was to find the best-fitting λ (regularization) parameter that would minimize mean squared error, by using nine variations of these resampling methods. For each of the nine resampling variations, 1,000 runs were performed to see how often a good fit, average fit, and poor fit λ value would be chosen. The resampling method that chose good fit values the greatest number of times was deemed the best method. Based on the results of the investigation, three general recommendations are made: (1) repeated k-fold cross-validation is the best method to select as a general-purpose resampling method; (2) k = 10 folds is a good choice in k-fold cross-validation; (3) Monte Carlo and bootstrap are underperformers, so they are not recommended as general-purpose resampling methods. At the same time, no resampling method was found to be uniformly better than the others.","PeriodicalId":46139,"journal":{"name":"Journal of Intelligent Systems","volume":"6 1","pages":""},"PeriodicalIF":3.0,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78871554","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}
Abstract Image data play important role in various real-time online and offline applications. Biomedical field has adopted the imaging system to detect, diagnose, and prevent several types of diseases and abnormalities. The biomedical imaging data contain huge information which requires huge storage space. Moreover, currently telemedicine and IoT based remote health monitoring systems are widely developed where data is transmitted from one place to another. Transmission of this type of huge data consumes more bandwidth. Along with this, during this transmission, the attackers can attack the communication channel and obtain the important and secret information. Hence, biomedical image compression and encryption are considered the solution to deal with these issues. Several techniques have been presented but achieving desired performance for combined module is a challenging task. Hence, in this work, a novel combined approach for image compression and encryption is developed. First, image compression scheme using wavelet transform is presented and later a cryptography scheme is presented using confusion and diffusion schemes. The outcome of the proposed approach is compared with various existing techniques. The experimental analysis shows that the proposed approach achieves better performance in terms of autocorrelation, histogram, information entropy, PSNR, MSE, and SSIM.
{"title":"HWCD: A hybrid approach for image compression using wavelet, encryption using confusion, and decryption using diffusion scheme","authors":"H. R. Latha, Alagarswamy Ramaprasath","doi":"10.1515/jisys-2022-9056","DOIUrl":"https://doi.org/10.1515/jisys-2022-9056","url":null,"abstract":"Abstract Image data play important role in various real-time online and offline applications. Biomedical field has adopted the imaging system to detect, diagnose, and prevent several types of diseases and abnormalities. The biomedical imaging data contain huge information which requires huge storage space. Moreover, currently telemedicine and IoT based remote health monitoring systems are widely developed where data is transmitted from one place to another. Transmission of this type of huge data consumes more bandwidth. Along with this, during this transmission, the attackers can attack the communication channel and obtain the important and secret information. Hence, biomedical image compression and encryption are considered the solution to deal with these issues. Several techniques have been presented but achieving desired performance for combined module is a challenging task. Hence, in this work, a novel combined approach for image compression and encryption is developed. First, image compression scheme using wavelet transform is presented and later a cryptography scheme is presented using confusion and diffusion schemes. The outcome of the proposed approach is compared with various existing techniques. The experimental analysis shows that the proposed approach achieves better performance in terms of autocorrelation, histogram, information entropy, PSNR, MSE, and SSIM.","PeriodicalId":46139,"journal":{"name":"Journal of Intelligent Systems","volume":"37 1","pages":""},"PeriodicalIF":3.0,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78967186","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}
Abstract Translation of long sentences in English is a complex problem in machine translation. This work briefly introduced the basic framework of intelligent machine translation algorithm and improved the long short-term memory (LSTM)-based intelligent machine translation algorithm by introducing the long sentence segmentation module and reordering module. Simulation experiments were conducted using the public corpus and the local corpus containing self-collected linguistic data. The improved algorithm was compared with machine translation algorithms based on a recurrent neural network and LSTM. The results suggested that the LSTM-based machine translation algorithm added with the long sentence segmentation module and reordering module effectively segmented long sentences and translated long English sentences more accurately, and the translation was more grammatically correct.
{"title":"An intelligent algorithm for fast machine translation of long English sentences","authors":"Hengheng He","doi":"10.1515/jisys-2022-0257","DOIUrl":"https://doi.org/10.1515/jisys-2022-0257","url":null,"abstract":"Abstract Translation of long sentences in English is a complex problem in machine translation. This work briefly introduced the basic framework of intelligent machine translation algorithm and improved the long short-term memory (LSTM)-based intelligent machine translation algorithm by introducing the long sentence segmentation module and reordering module. Simulation experiments were conducted using the public corpus and the local corpus containing self-collected linguistic data. The improved algorithm was compared with machine translation algorithms based on a recurrent neural network and LSTM. The results suggested that the LSTM-based machine translation algorithm added with the long sentence segmentation module and reordering module effectively segmented long sentences and translated long English sentences more accurately, and the translation was more grammatically correct.","PeriodicalId":46139,"journal":{"name":"Journal of Intelligent Systems","volume":"22 1","pages":""},"PeriodicalIF":3.0,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83202682","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}
Abstract Remote sensing image technology to the ground has important guiding significance in disaster assessment and emergency rescue deployment. In order to realize the fast automatic registration of multi-sensor remote sensing images, the remote sensing image block registration idea is introduced, and the image reconstruction is processed by using the conjugate gradient descent (CGD) method. The scale-invariant feature transformation (SIFT) algorithm is improved and optimized by combining the function-fitting method. By this way, it can improve the registration accuracy and efficiency of multi-sensor remote sensing images. The results show that the average peak signal-to-noise ratio of the image processed by the CGD method is 25.428. The average root mean square value is 17.442. The average image processing time is 6.093 s. These indicators are better than the passive filter algorithm and the gradient descent method. The average accuracy of image registration of the improved SIFT registration method is 96.37%, and the average image registration time is 2.14 s. These indicators are significantly better than the traditional SIFT algorithm and speeded-up robust features algorithm. It is proved that the improved SIFT registration method can effectively improve the accuracy and operation efficiency of multi-sensor remote sensing image registration methods. The improved SIFT registration method effectively solves the problems of low accuracy and long time consumption of traditional multi-sensor remote sensing image fast registration methods. While maintaining high registration accuracy, it improves the image registration speed and provides technical support for a rapid disaster assessment after major disasters such as earthquakes and floods. And it has an important value for the development of the efficient post-disaster rescue deployment.
{"title":"Multi-sensor remote sensing image alignment based on fast algorithms","authors":"Tao Shu","doi":"10.1515/jisys-2022-0289","DOIUrl":"https://doi.org/10.1515/jisys-2022-0289","url":null,"abstract":"Abstract Remote sensing image technology to the ground has important guiding significance in disaster assessment and emergency rescue deployment. In order to realize the fast automatic registration of multi-sensor remote sensing images, the remote sensing image block registration idea is introduced, and the image reconstruction is processed by using the conjugate gradient descent (CGD) method. The scale-invariant feature transformation (SIFT) algorithm is improved and optimized by combining the function-fitting method. By this way, it can improve the registration accuracy and efficiency of multi-sensor remote sensing images. The results show that the average peak signal-to-noise ratio of the image processed by the CGD method is 25.428. The average root mean square value is 17.442. The average image processing time is 6.093 s. These indicators are better than the passive filter algorithm and the gradient descent method. The average accuracy of image registration of the improved SIFT registration method is 96.37%, and the average image registration time is 2.14 s. These indicators are significantly better than the traditional SIFT algorithm and speeded-up robust features algorithm. It is proved that the improved SIFT registration method can effectively improve the accuracy and operation efficiency of multi-sensor remote sensing image registration methods. The improved SIFT registration method effectively solves the problems of low accuracy and long time consumption of traditional multi-sensor remote sensing image fast registration methods. While maintaining high registration accuracy, it improves the image registration speed and provides technical support for a rapid disaster assessment after major disasters such as earthquakes and floods. And it has an important value for the development of the efficient post-disaster rescue deployment.","PeriodicalId":46139,"journal":{"name":"Journal of Intelligent Systems","volume":"25 1","pages":""},"PeriodicalIF":3.0,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82616252","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}
Abstract Aiming at the problem that users’ check-in interest preferences in social networks have complex time dependences, which leads to inaccurate point-of-interest (POI) recommendations, a location-based POI recommendation model using deep learning for social network big data is proposed. First, the original data are fed into an embedding layer of the model for dense vector representation and to obtain the user’s check-in sequence (UCS) and space-time interval information. Then, the UCS and spatiotemporal interval information are sent into a bidirectional long-term memory model for detailed analysis, where the UCS and location sequence representation are updated using a self-attention mechanism. Finally, candidate POIs are compared with the user’s preferences, and a POI sequence with three consecutive recommended locations is generated. The experimental analysis shows that the model performs best when the Huber loss function is used and the number of training iterations is set to 200. In the Foursquare dataset, Recall@20 and NDCG@20 reach 0.418 and 0.143, and in the Gowalla dataset, the corresponding values are 0.387 and 0.148.
{"title":"A BiLSTM-attention-based point-of-interest recommendation algorithm","authors":"Aichuan Li, Fuzhi Liu","doi":"10.1515/jisys-2023-0033","DOIUrl":"https://doi.org/10.1515/jisys-2023-0033","url":null,"abstract":"Abstract Aiming at the problem that users’ check-in interest preferences in social networks have complex time dependences, which leads to inaccurate point-of-interest (POI) recommendations, a location-based POI recommendation model using deep learning for social network big data is proposed. First, the original data are fed into an embedding layer of the model for dense vector representation and to obtain the user’s check-in sequence (UCS) and space-time interval information. Then, the UCS and spatiotemporal interval information are sent into a bidirectional long-term memory model for detailed analysis, where the UCS and location sequence representation are updated using a self-attention mechanism. Finally, candidate POIs are compared with the user’s preferences, and a POI sequence with three consecutive recommended locations is generated. The experimental analysis shows that the model performs best when the Huber loss function is used and the number of training iterations is set to 200. In the Foursquare dataset, Recall@20 and NDCG@20 reach 0.418 and 0.143, and in the Gowalla dataset, the corresponding values are 0.387 and 0.148.","PeriodicalId":46139,"journal":{"name":"Journal of Intelligent Systems","volume":"60 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135104735","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}
Abstract Waste classification is the issue of sorting rubbish into valuable categories for efficient waste management. Problems arise from issues such as individual ignorance or inactivity and more overt issues like pollution in the environment, lack of resources, or a malfunctioning system. Education, established behaviors, an improved infrastructure, technology, and legislative incentives to promote effective trash sorting and management are all necessary for a solution to be implemented. For solid waste management and recycling efforts to be successful, waste materials must be sorted appropriately. This study evaluates the effectiveness of several deep learning (DL) models for the challenge of waste material classification. The focus will be on finding the best DL technique for solid waste classification. This study extensively compares several DL architectures (Resnet50, GoogleNet, InceptionV3, and Xception). Images of various types of trash are amassed and cleaned up to form a dataset. Accuracy, precision, recall, and F 1 score are only a few measures used to assess the performance of the many DL models trained and tested on this dataset. ResNet50 showed impressive performance in waste material classification, with 95% accuracy, 95.4% precision, 95% recall, and 94.8% in the F 1 score, with only two incorrect categories in the glass class. All classes are correctly classified with an F 1 score of 100% due to Inception V3’s remarkable accuracy, precision, recall, and F 1 score. Xception’s classification accuracy was excellent (100%), with a few difficulties in the glass and trash categories. With a good 90.78% precision, 100% recall, and 89.81% F 1 score, GoogleNet performed admirably. This study highlights the significance of using models based on DL for categorizing trash. The results open the way for enhanced trash sorting and recycling operations, contributing to an economically and ecologically friendly future.
{"title":"Waste material classification using performance evaluation of deep learning models","authors":"Israa Badr Al-Mashhadani","doi":"10.1515/jisys-2023-0064","DOIUrl":"https://doi.org/10.1515/jisys-2023-0064","url":null,"abstract":"Abstract Waste classification is the issue of sorting rubbish into valuable categories for efficient waste management. Problems arise from issues such as individual ignorance or inactivity and more overt issues like pollution in the environment, lack of resources, or a malfunctioning system. Education, established behaviors, an improved infrastructure, technology, and legislative incentives to promote effective trash sorting and management are all necessary for a solution to be implemented. For solid waste management and recycling efforts to be successful, waste materials must be sorted appropriately. This study evaluates the effectiveness of several deep learning (DL) models for the challenge of waste material classification. The focus will be on finding the best DL technique for solid waste classification. This study extensively compares several DL architectures (Resnet50, GoogleNet, InceptionV3, and Xception). Images of various types of trash are amassed and cleaned up to form a dataset. Accuracy, precision, recall, and F 1 score are only a few measures used to assess the performance of the many DL models trained and tested on this dataset. ResNet50 showed impressive performance in waste material classification, with 95% accuracy, 95.4% precision, 95% recall, and 94.8% in the F 1 score, with only two incorrect categories in the glass class. All classes are correctly classified with an F 1 score of 100% due to Inception V3’s remarkable accuracy, precision, recall, and F 1 score. Xception’s classification accuracy was excellent (100%), with a few difficulties in the glass and trash categories. With a good 90.78% precision, 100% recall, and 89.81% F 1 score, GoogleNet performed admirably. This study highlights the significance of using models based on DL for categorizing trash. The results open the way for enhanced trash sorting and recycling operations, contributing to an economically and ecologically friendly future.","PeriodicalId":46139,"journal":{"name":"Journal of Intelligent Systems","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135561223","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：