Pub Date : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941281
Yifan Zhang, Qian Shen, Jian Guo, Yiwen Jia
In order to study how to construct a suitable portfolio trading strategy of traditional financial products and new kinds of financial products to help investors avoid risks and obtain more returns, we use pair trading models, polynomial regression models, and a machine learning-based combined model we designed to make a simulated trading. In the simulation of gold and bitcoin trading, our combined model achieved better results and avoided the shortcomings of the pair trading model and the polynomial regression model. We suggest that investors add constraints to the combined model according to the actual situation of financial products, and use it to forecast and make decisions on portfolio tradings.
{"title":"Portfolio Trading of Financial Products Based on Machine Learning","authors":"Yifan Zhang, Qian Shen, Jian Guo, Yiwen Jia","doi":"10.1109/ICMLC56445.2022.9941281","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941281","url":null,"abstract":"In order to study how to construct a suitable portfolio trading strategy of traditional financial products and new kinds of financial products to help investors avoid risks and obtain more returns, we use pair trading models, polynomial regression models, and a machine learning-based combined model we designed to make a simulated trading. In the simulation of gold and bitcoin trading, our combined model achieved better results and avoided the shortcomings of the pair trading model and the polynomial regression model. We suggest that investors add constraints to the combined model according to the actual situation of financial products, and use it to forecast and make decisions on portfolio tradings.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121188314","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941330
Yulan Luo, Yi Ji
Volatility in the stock market has a significant impact on all finance-related fields. As an important part of stock data, the adjusted closing price often reflects the attention of market funds to a stock, helping predict the market movement of the next trading day, especially for short-term investors. With the development of artificial intelligence technology, the machine learning algorithms are widely applied to predict stock trends. However, the noisy, nonlinear, and chaotic nature of stock price changes makes the prediction not accurate enough. Hence, we proposed a hybrid prediction model combining improved particle swarm optimization (IPSO) and long short-term memory (LSTM) neural network to predict the adjusted closing price of the stock. In this paper, nonlinear methods are presented to optimize the velocity inertia weight and learning factors of traditional particle swarm optimization (PSO). Meanwhile, IPSO is used to optimize the hyperparameters of LSTM neural network to improve its prediction accuracy. The experiments proved that the proposed IPSO-LSTM outperformed the Autoregressive Integrated Moving Average model (ARIMA), LSTM, and PSO-LSTM on the prediction of the S&P 500 Index. Furthermore, the Dow Jones Industrial Average Index (DJI) and Nasdaq Composite Index (IXIC) were chosen to verify the accuracy and robustness of the model we put forward.
{"title":"Prediction of the Stock Adjusted Closing Price Based On Improved PSO-LSTM Neural Network","authors":"Yulan Luo, Yi Ji","doi":"10.1109/ICMLC56445.2022.9941330","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941330","url":null,"abstract":"Volatility in the stock market has a significant impact on all finance-related fields. As an important part of stock data, the adjusted closing price often reflects the attention of market funds to a stock, helping predict the market movement of the next trading day, especially for short-term investors. With the development of artificial intelligence technology, the machine learning algorithms are widely applied to predict stock trends. However, the noisy, nonlinear, and chaotic nature of stock price changes makes the prediction not accurate enough. Hence, we proposed a hybrid prediction model combining improved particle swarm optimization (IPSO) and long short-term memory (LSTM) neural network to predict the adjusted closing price of the stock. In this paper, nonlinear methods are presented to optimize the velocity inertia weight and learning factors of traditional particle swarm optimization (PSO). Meanwhile, IPSO is used to optimize the hyperparameters of LSTM neural network to improve its prediction accuracy. The experiments proved that the proposed IPSO-LSTM outperformed the Autoregressive Integrated Moving Average model (ARIMA), LSTM, and PSO-LSTM on the prediction of the S&P 500 Index. Furthermore, the Dow Jones Industrial Average Index (DJI) and Nasdaq Composite Index (IXIC) were chosen to verify the accuracy and robustness of the model we put forward.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"02 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130590038","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941291
Zhongkai Zhao, Tatsuhito Hasegawa
Transfer learning improves problem-solving efficiency by transferring the learned knowledge from the source domain to the target domain. In transfer learning, using a large amount of data for pre-training is beneficial to improve the robustness of the model. Data differ significantly when the domain changes in Sensor-Based human activity recognition (HAR). Currently, in HAR, data usage is relatively independent, lacking source domains with massive data and rich labels. This paper proposes a new pre-training method using multiple domain datasets to construct a domain-robust pre-training model. We divide the pre-training dataset into basic and complex activities scenarios by considering the difference in activity classification. We evaluate the classification scenarios that are most beneficial for sensor-based HAR based on the constituted dataset and using deep convolutional networks. We show that our method verified the influence of the source domain on transfer learning in sensor-based HAR. By constructing a sizeable correlated source domain, our method can enhance the generalization ability of the network model. This paper also demonstrated that large-scale and basic activity classification datasets can be better used as pre-training models to participate in HAR classification tasks.
{"title":"Domain-Robust Pre-Training Method for the Sensor-Based Human Activity Recognition","authors":"Zhongkai Zhao, Tatsuhito Hasegawa","doi":"10.1109/ICMLC56445.2022.9941291","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941291","url":null,"abstract":"Transfer learning improves problem-solving efficiency by transferring the learned knowledge from the source domain to the target domain. In transfer learning, using a large amount of data for pre-training is beneficial to improve the robustness of the model. Data differ significantly when the domain changes in Sensor-Based human activity recognition (HAR). Currently, in HAR, data usage is relatively independent, lacking source domains with massive data and rich labels. This paper proposes a new pre-training method using multiple domain datasets to construct a domain-robust pre-training model. We divide the pre-training dataset into basic and complex activities scenarios by considering the difference in activity classification. We evaluate the classification scenarios that are most beneficial for sensor-based HAR based on the constituted dataset and using deep convolutional networks. We show that our method verified the influence of the source domain on transfer learning in sensor-based HAR. By constructing a sizeable correlated source domain, our method can enhance the generalization ability of the network model. This paper also demonstrated that large-scale and basic activity classification datasets can be better used as pre-training models to participate in HAR classification tasks.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"200 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133724422","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}
In Japan, drunk driving is prohibited under the Road Traffic Law, and penalties are set at a breath alcohol concentration of 0.15 mg/l. However, 300 cases of drunk driving occur annually even when the breath alcohol concentration is below the standard value. This suggests that even small amounts of alcohol consumption may cause a decline in brain function. In this study, we evaluated the brain function caused by low-intensity drinking using event-related potentials, a type of electroencephalogram (EEG). The results showed that breath alcohol concentration increased significantly (p < 0.05) at 10, 30, and 50 minutes after drinking compared to before drinking. Event-related potentials during these time periods also changed significantly (p < 0.05). On the other hand, there was no significant difference in expiratory alcohol concentration during the first 70 minutes after drinking, but there was a significant change in event-related potentials. The present study suggests that low alcohol intake at low loads causes a decrease in brain function.
{"title":"A Study of Breath Alcohol Concentration Fluctuations and Cognitive Decline Due to Low-Impact Drinking","authors":"Yuichi Sato, Kosuke Nagano, Fumiya Kinoshita, Hideaki Touyama","doi":"10.1109/ICMLC56445.2022.9941327","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941327","url":null,"abstract":"In Japan, drunk driving is prohibited under the Road Traffic Law, and penalties are set at a breath alcohol concentration of 0.15 mg/l. However, 300 cases of drunk driving occur annually even when the breath alcohol concentration is below the standard value. This suggests that even small amounts of alcohol consumption may cause a decline in brain function. In this study, we evaluated the brain function caused by low-intensity drinking using event-related potentials, a type of electroencephalogram (EEG). The results showed that breath alcohol concentration increased significantly (p < 0.05) at 10, 30, and 50 minutes after drinking compared to before drinking. Event-related potentials during these time periods also changed significantly (p < 0.05). On the other hand, there was no significant difference in expiratory alcohol concentration during the first 70 minutes after drinking, but there was a significant change in event-related potentials. The present study suggests that low alcohol intake at low loads causes a decrease in brain function.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133811808","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941326
Shodai Ito, Noboru Takagi, K. Sawai, H. Masuta, T. Motoyoshi
Much research has been done on pattern recognition in line drawings. Converting raster graphics into vector graphics is one such examples. Vector graphics are composed of meaningful basic components such as lines, curves, and parabolas etc. However, converting raster graphic to a vector graphic is difficult because the structures of the basic components must be recognized. Therefore, we propose a semantic segmentation method for converting line drawings in raster format into vector format and verify the accuracy of the extraction of basic components and the processing time through computer experiments.
{"title":"Fast Semantic Segmentation for Vectorization of Line Drawings Based on Deep Neural Networks","authors":"Shodai Ito, Noboru Takagi, K. Sawai, H. Masuta, T. Motoyoshi","doi":"10.1109/ICMLC56445.2022.9941326","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941326","url":null,"abstract":"Much research has been done on pattern recognition in line drawings. Converting raster graphics into vector graphics is one such examples. Vector graphics are composed of meaningful basic components such as lines, curves, and parabolas etc. However, converting raster graphic to a vector graphic is difficult because the structures of the basic components must be recognized. Therefore, we propose a semantic segmentation method for converting line drawings in raster format into vector format and verify the accuracy of the extraction of basic components and the processing time through computer experiments.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114231080","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941303
Farchan Hakim Raswa, Indra Yusuf Kinarta, Reza Pulungan, A. Harjoko, Chung-Ting Lee, Yung-Hui Li, Jia-Ching Wang
Fingerprint has a competent level of uniqueness because the various features can form different patterns in humans. It is a verification requirement in various aspects, such as mobile phone, banking accounts, attendance, etc. One of the preventive measures in maintaining performance is liveness detection. We deep exploited the handcrafted method to achieve adequate performance. To encapsulate the noise possibility, we added the Bayes shrink-wavelet transform as the noise removal. So, the noise obtained in the fingerprint image can be minimized but keep the quality of the fingerprint image is in good condition. Then, we conjugated the spatial and frequency domain in pixel neighborhood distribution using the local binary pattern (LBP) and local phase quantization (LPQ) feature. Finally, we mapped the learning stage using a prominent classifier, i.e., a support vector machine (SVM). Our experiment was evaluated with LivDet 2015 dataset. The proposed method has achieved sustainable results regarding average error rate (AER).
{"title":"Fingerprint Liveness Detection Using Denoised-Bayes Shrink Wavelet and Aggregated Local Spatial and Frequency Features","authors":"Farchan Hakim Raswa, Indra Yusuf Kinarta, Reza Pulungan, A. Harjoko, Chung-Ting Lee, Yung-Hui Li, Jia-Ching Wang","doi":"10.1109/ICMLC56445.2022.9941303","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941303","url":null,"abstract":"Fingerprint has a competent level of uniqueness because the various features can form different patterns in humans. It is a verification requirement in various aspects, such as mobile phone, banking accounts, attendance, etc. One of the preventive measures in maintaining performance is liveness detection. We deep exploited the handcrafted method to achieve adequate performance. To encapsulate the noise possibility, we added the Bayes shrink-wavelet transform as the noise removal. So, the noise obtained in the fingerprint image can be minimized but keep the quality of the fingerprint image is in good condition. Then, we conjugated the spatial and frequency domain in pixel neighborhood distribution using the local binary pattern (LBP) and local phase quantization (LPQ) feature. Finally, we mapped the learning stage using a prominent classifier, i.e., a support vector machine (SVM). Our experiment was evaluated with LivDet 2015 dataset. The proposed method has achieved sustainable results regarding average error rate (AER).","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116327576","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941299
Chun-Ming Tsai, F. Shih, J. Hsieh
In order to improve the driving safety and reduce traffic congestion during holidays and work hours, a real-time vehicle detection and counting system is a very urgently needed system. In this paper, a lane-based vehicle counting system using deep-learning networks is proposed. Our method includes YOLO vehicle detection and lane-based vehicle counting. From the vehicle detection experimental results, YOLOv3-spp has the highest Precision, Recall, and F1 score, which achieve all 100% among three YOLOv3 methods and two YOLOv2 methods. From the vehicle counting experimental results, YOLOv3-608 has the highest Accuracy, Precision and F1 scores, which achieve 91.4%, 99.3%, and 95.3% among three YOLOv3 methods, two YOLOv2 methods, and one SSD method.
{"title":"Real-Time Vehicle Counting by Deep-Learning Networks","authors":"Chun-Ming Tsai, F. Shih, J. Hsieh","doi":"10.1109/ICMLC56445.2022.9941299","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941299","url":null,"abstract":"In order to improve the driving safety and reduce traffic congestion during holidays and work hours, a real-time vehicle detection and counting system is a very urgently needed system. In this paper, a lane-based vehicle counting system using deep-learning networks is proposed. Our method includes YOLO vehicle detection and lane-based vehicle counting. From the vehicle detection experimental results, YOLOv3-spp has the highest Precision, Recall, and F1 score, which achieve all 100% among three YOLOv3 methods and two YOLOv2 methods. From the vehicle counting experimental results, YOLOv3-608 has the highest Accuracy, Precision and F1 scores, which achieve 91.4%, 99.3%, and 95.3% among three YOLOv3 methods, two YOLOv2 methods, and one SSD method.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"187 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114969950","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941334
Koki Takenaka, Tatsuhito Hasegawa
Deep learning methods contribute to improve the estimation accuracy in human activity recognition (HAR) using sensor data. In general, the dataset used in HAR consists of accelerometer data and activity labels. Because of the widespread use of mobile devices, large amount of accelerometer sensor data without activity labels can be easily collected. The problem of annotation needs a large amount of time-consuming cost and human labor to annotate a activity labels to recorded sensor data. Therefore, we need a method to make deep learning models acquire feature representations from accelerometer data without activity labels in HAR. In this study, based on the unsupervised representation learning method proposed in image recognition, we proposed a new unsupervised representation learning method which combines segment discrimination (SD), autoencoder (AE) and feature independent softmax (FIS). Our experimental results showed that our proposed method outperformed the conventional method in fine-tuning accuracy in HAR.
{"title":"Unsupervised Representation Learning Method In Sensor Based Human Activity Recognition","authors":"Koki Takenaka, Tatsuhito Hasegawa","doi":"10.1109/ICMLC56445.2022.9941334","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941334","url":null,"abstract":"Deep learning methods contribute to improve the estimation accuracy in human activity recognition (HAR) using sensor data. In general, the dataset used in HAR consists of accelerometer data and activity labels. Because of the widespread use of mobile devices, large amount of accelerometer sensor data without activity labels can be easily collected. The problem of annotation needs a large amount of time-consuming cost and human labor to annotate a activity labels to recorded sensor data. Therefore, we need a method to make deep learning models acquire feature representations from accelerometer data without activity labels in HAR. In this study, based on the unsupervised representation learning method proposed in image recognition, we proposed a new unsupervised representation learning method which combines segment discrimination (SD), autoencoder (AE) and feature independent softmax (FIS). Our experimental results showed that our proposed method outperformed the conventional method in fine-tuning accuracy in HAR.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115180824","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941325
Rung-Ching Chen, William Eric Manongga, Christine Dewi, Hung-Yi Chen
Sign language is challenging to understand and needs a lot of practice before it can be mastered. With the growth of the deaf and the hard-of-hearing community, researchers are trying to find an effective way to understand sign language. This study will utilize hand landmarks to detect digits in sign language. Three models trained with different features will be used to compare their accuracy. The first model will be trained using the hand images only, the second model will be trained using the hand image and the hand landmarks, and the third model will be trained using the hand landmarks only. Mediapipe will be used to extract the hand landmark features, which is one of the features used by the model. The study results show that the first and second models have better training and testing accuracy than the third. However, the third model is superior when evaluated using the validation dataset with 85% accuracy, compared to 23.30% and 41.70% for the first and second models.
{"title":"Automatic Digit Hand Sign Detection With Hand Landmark","authors":"Rung-Ching Chen, William Eric Manongga, Christine Dewi, Hung-Yi Chen","doi":"10.1109/ICMLC56445.2022.9941325","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941325","url":null,"abstract":"Sign language is challenging to understand and needs a lot of practice before it can be mastered. With the growth of the deaf and the hard-of-hearing community, researchers are trying to find an effective way to understand sign language. This study will utilize hand landmarks to detect digits in sign language. Three models trained with different features will be used to compare their accuracy. The first model will be trained using the hand images only, the second model will be trained using the hand image and the hand landmarks, and the third model will be trained using the hand landmarks only. Mediapipe will be used to extract the hand landmark features, which is one of the features used by the model. The study results show that the first and second models have better training and testing accuracy than the third. However, the third model is superior when evaluated using the validation dataset with 85% accuracy, compared to 23.30% and 41.70% for the first and second models.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132835249","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941296
R. Chen, Wanjun Yang, Kuei-Chien Chiu
Predicting stock prices has always been an attractive issue. Past literature has focused on the impact of historical stock prices and social media sentiment on stock prices, ignoring the impact on the three major corporations that account for most stock transactions. In this paper, we add the three significant corporations as the dataset in the stock trading price, but the corporate trading data announced by the stock exchange has only been available since May 2012, so the data sample is less than ten years. In the target dataset, we compared the model with the ARIMA and LSTM for error, and the migration learning model outperformed the other two models.
{"title":"Transfer Learning and LSTM to Predict Stock Price","authors":"R. Chen, Wanjun Yang, Kuei-Chien Chiu","doi":"10.1109/ICMLC56445.2022.9941296","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941296","url":null,"abstract":"Predicting stock prices has always been an attractive issue. Past literature has focused on the impact of historical stock prices and social media sentiment on stock prices, ignoring the impact on the three major corporations that account for most stock transactions. In this paper, we add the three significant corporations as the dataset in the stock trading price, but the corporate trading data announced by the stock exchange has only been available since May 2012, so the data sample is less than ten years. In the target dataset, we compared the model with the ARIMA and LSTM for error, and the migration learning model outperformed the other two models.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"09 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127311307","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：