Pub Date : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00058
Kin-Hon Ho, Wai-Han Chiu, Chin Li
We conduct a network analysis with centrality measures, using historical daily close prices of top 120 cryptocurrencies between 2013 and 2020, to study and understand the dynamic evolution and characteristics of the cryptocurrency market. Our study has three primary findings: (1) the overall cross-return correlation among the cryptocurrencies is weakening from 2013 to 2016 and then strengthening thereafter; (2) cryptocurrencies that are primarily used for transaction payment, notably BTC, dominate the market until mid-2016, followed by those developed for applications using blockchain as the underlying technology, particularly data storage and recording such as MAID and FCT, between mid-2016 and mid-2017. Since then, ETH, alongside with its strongly correlated cryptocurrencies have replaced BTC to become the benchmark cryptocurrencies. Furthermore, during the outbreak of COVID-19, QTUM and BNB have intermittently replaced ETH to take the leading positions due to their active community engagement during the pandemic; (3) centrality measures are useful features in improving the prediction accuracy of the short-term cryptocurrency price movement.
{"title":"A Short-Term Cryptocurrency Price Movement Prediction Using Centrality Measures","authors":"Kin-Hon Ho, Wai-Han Chiu, Chin Li","doi":"10.1109/ICDMW51313.2020.00058","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00058","url":null,"abstract":"We conduct a network analysis with centrality measures, using historical daily close prices of top 120 cryptocurrencies between 2013 and 2020, to study and understand the dynamic evolution and characteristics of the cryptocurrency market. Our study has three primary findings: (1) the overall cross-return correlation among the cryptocurrencies is weakening from 2013 to 2016 and then strengthening thereafter; (2) cryptocurrencies that are primarily used for transaction payment, notably BTC, dominate the market until mid-2016, followed by those developed for applications using blockchain as the underlying technology, particularly data storage and recording such as MAID and FCT, between mid-2016 and mid-2017. Since then, ETH, alongside with its strongly correlated cryptocurrencies have replaced BTC to become the benchmark cryptocurrencies. Furthermore, during the outbreak of COVID-19, QTUM and BNB have intermittently replaced ETH to take the leading positions due to their active community engagement during the pandemic; (3) centrality measures are useful features in improving the prediction accuracy of the short-term cryptocurrency price movement.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114792531","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00019
L. Subhashini, Yuefeng Li, Jinglan Zhang, Ajantha S Atukorale
The problem of uncertainty is a challenging issue to solve in opinion mining models. Existing models that use machine learning algorithms are unable to identify uncertainty within online customer reviews because of broad uncertain boundaries. Many researchers have developed fuzzy models to solve this problem. However, the problem of large uncertain boundaries remains with fuzzy models. The common challenging issue is that there is a big uncertain boundary between positive and negative classes as user reviews (or opinions) include many uncertainties. Dealing with these uncertainties is problematic due in many frequently used words may be non-relevant. This paper proposes a three-way based framework which integrates fuzzy concepts and deep learning together to solve the problem of uncertainty. Many experiments were conducted using movie review and ebook review datasets. The experimental results show that the proposed three-way framework is useful for dealing with uncertainties in opinions and we were able to show that significant F-measure for two benchmark dataset.
{"title":"Integration of Fuzzy and Deep Learning in Three-Way Decisions","authors":"L. Subhashini, Yuefeng Li, Jinglan Zhang, Ajantha S Atukorale","doi":"10.1109/ICDMW51313.2020.00019","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00019","url":null,"abstract":"The problem of uncertainty is a challenging issue to solve in opinion mining models. Existing models that use machine learning algorithms are unable to identify uncertainty within online customer reviews because of broad uncertain boundaries. Many researchers have developed fuzzy models to solve this problem. However, the problem of large uncertain boundaries remains with fuzzy models. The common challenging issue is that there is a big uncertain boundary between positive and negative classes as user reviews (or opinions) include many uncertainties. Dealing with these uncertainties is problematic due in many frequently used words may be non-relevant. This paper proposes a three-way based framework which integrates fuzzy concepts and deep learning together to solve the problem of uncertainty. Many experiments were conducted using movie review and ebook review datasets. The experimental results show that the proposed three-way framework is useful for dealing with uncertainties in opinions and we were able to show that significant F-measure for two benchmark dataset.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114267495","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00045
JunYong Tong, Nick Torenvliet
This paper proposes a streaming anomaly detection algorithm using variational Bayesian non-parametric methods. We extend the use of Dirichlet process mixture models to anomaly detection for online streaming data through the use of streaming variational bayes method and a cohesion function. Using our algorithm, we were able to update model parameters sequentially near real-time, using a fixed amount of computational resources. The algorithm was able to capture the temporal dynamics of the data and enabled good online anomaly detection. We demonstrate the performance, and discuss results, of the algorithm on an industrial datasets with anomalies provided by a local utility.
{"title":"Temporally-Reweighted Dirichlet Process Mixture Anomaly Detector","authors":"JunYong Tong, Nick Torenvliet","doi":"10.1109/ICDMW51313.2020.00045","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00045","url":null,"abstract":"This paper proposes a streaming anomaly detection algorithm using variational Bayesian non-parametric methods. We extend the use of Dirichlet process mixture models to anomaly detection for online streaming data through the use of streaming variational bayes method and a cohesion function. Using our algorithm, we were able to update model parameters sequentially near real-time, using a fixed amount of computational resources. The algorithm was able to capture the temporal dynamics of the data and enabled good online anomaly detection. We demonstrate the performance, and discuss results, of the algorithm on an industrial datasets with anomalies provided by a local utility.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130792155","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00057
Geet Shingi
The number of defaults in bank loans have recently been increasing in the past years. However, the process of sanctioning the loan has still been done manually in many of the banking organizations. Dependency on human intervention and delay in results have been the biggest obstacles in this system. While implementing machine learning models for banking applications, the security of sensitive customer banking data has always been a crucial concern and with strong legislative rules in place, sharing of data with other organizations is not possible. Along with this, the loan dataset is highly imbalanced, there are very few samples of defaults as compared to repaid loans. Hence, these problems make the default prediction system difficult to learn the patterns of defaults and thus difficult to predict them. Previous machine learning-based approaches to automate the process have been training models on the same organization's data but in today's world, classifying the loan application on the data within the organizations is no longer sufficient and a feasible solution. In this paper, we propose a federated learning-based approach for the prediction of loan applications that are less likely to be repaid which helps in resolving the above mentioned issues by sharing the weight of the model which are aggregated at the central server. The federated system is coupled with Synthetic Minority Over-sampling Technique(SMOTE) to solve the problem of imbalanced training data. Further, The federated system is coupled with a weighted aggregation based on the number of samples and performance of a worker on his dataset to further augment the performance. The improved performance by our model on publicly available real-world data further validates the same. Flexible, aggregated models can prove to be crucial in keeping out the defaulters in loan applications.
{"title":"A federated learning based approach for loan defaults prediction","authors":"Geet Shingi","doi":"10.1109/ICDMW51313.2020.00057","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00057","url":null,"abstract":"The number of defaults in bank loans have recently been increasing in the past years. However, the process of sanctioning the loan has still been done manually in many of the banking organizations. Dependency on human intervention and delay in results have been the biggest obstacles in this system. While implementing machine learning models for banking applications, the security of sensitive customer banking data has always been a crucial concern and with strong legislative rules in place, sharing of data with other organizations is not possible. Along with this, the loan dataset is highly imbalanced, there are very few samples of defaults as compared to repaid loans. Hence, these problems make the default prediction system difficult to learn the patterns of defaults and thus difficult to predict them. Previous machine learning-based approaches to automate the process have been training models on the same organization's data but in today's world, classifying the loan application on the data within the organizations is no longer sufficient and a feasible solution. In this paper, we propose a federated learning-based approach for the prediction of loan applications that are less likely to be repaid which helps in resolving the above mentioned issues by sharing the weight of the model which are aggregated at the central server. The federated system is coupled with Synthetic Minority Over-sampling Technique(SMOTE) to solve the problem of imbalanced training data. Further, The federated system is coupled with a weighted aggregation based on the number of samples and performance of a worker on his dataset to further augment the performance. The improved performance by our model on publicly available real-world data further validates the same. Flexible, aggregated models can prove to be crucial in keeping out the defaulters in loan applications.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"59 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123874255","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00119
Pierre-Antoine Laharotte, Romain Billot, Nour-Eddin El Faouzi
Can we expose the relationship between the physical dynamics of a network and its predictability? To contribute to this point, we propose a dimensionality reduction method for network states prediction based on spatiotemporal data. The method is intended to deal with large scale networks, where only a subset of critical links can be relevant for accurate multidimensional prediction (MIMO) performances. The algorithm is based on Latent Dirichlet Allocation (LDA) to highlight relevant topics in terms of networks dynamics. The feature selection trick relies on the assumption that the most representative links of the most dominant topics are critical links for short term prediction. The method is fully implemented to an original application field: short term road traffic prediction on large scale urban networks based on GPS data. Results highlight significant reductions in dimensionality and execution time, a global improvement of prediction performances as well as a better resilience to non recurrent traffic flow conditions.
{"title":"Detecting Dynamic Critical Links within Large Scale Network for Traffic State Prediction","authors":"Pierre-Antoine Laharotte, Romain Billot, Nour-Eddin El Faouzi","doi":"10.1109/ICDMW51313.2020.00119","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00119","url":null,"abstract":"Can we expose the relationship between the physical dynamics of a network and its predictability? To contribute to this point, we propose a dimensionality reduction method for network states prediction based on spatiotemporal data. The method is intended to deal with large scale networks, where only a subset of critical links can be relevant for accurate multidimensional prediction (MIMO) performances. The algorithm is based on Latent Dirichlet Allocation (LDA) to highlight relevant topics in terms of networks dynamics. The feature selection trick relies on the assumption that the most representative links of the most dominant topics are critical links for short term prediction. The method is fully implemented to an original application field: short term road traffic prediction on large scale urban networks based on GPS data. Results highlight significant reductions in dimensionality and execution time, a global improvement of prediction performances as well as a better resilience to non recurrent traffic flow conditions.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121991379","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00022
A. Ramkissoon, Shareeda Mohammed
The existence of fake news is a problem challenging today's social media enabled world. Fake news can be classified using varying methods. Predicting and detecting fake news has proven to be challenging even for machine learning algorithms. This research attempts to investigate nine such machine learning algorithms to understand their performance with Credibility Based Fake News Detection. This study uses a standard dataset with features relating to the credibility of news publishers. These features are analysed using each of these algorithms. The results of these experiments are analysed using four evaluation methodologies. The analysis reveals varying performance with the use of each of the nine methods. Based upon our selected dataset, one of these methods has proven to be most appropriate for the purpose of Credibility Based Fake News Detection.
{"title":"An Experimental Evaluation of Data Classification Models for Credibility Based Fake News Detection","authors":"A. Ramkissoon, Shareeda Mohammed","doi":"10.1109/ICDMW51313.2020.00022","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00022","url":null,"abstract":"The existence of fake news is a problem challenging today's social media enabled world. Fake news can be classified using varying methods. Predicting and detecting fake news has proven to be challenging even for machine learning algorithms. This research attempts to investigate nine such machine learning algorithms to understand their performance with Credibility Based Fake News Detection. This study uses a standard dataset with features relating to the credibility of news publishers. These features are analysed using each of these algorithms. The results of these experiments are analysed using four evaluation methodologies. The analysis reveals varying performance with the use of each of the nine methods. Based upon our selected dataset, one of these methods has proven to be most appropriate for the purpose of Credibility Based Fake News Detection.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"100 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121053428","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00127
Y. Kostyuchenko, Qingshan Jiang
The globalization of the pharmaceutical supply chain has lead to new challenges, the leading position among them is the fight against falsified and substandard pharmaceutical products. Such kind of products causes ineffective or harmful therapies all over the world. Traditional centralized technical tools can hardly satisfy the requirements of the changing industry. In this paper, we research the application of Blockchain solutions to modernize the drug supply chain and minimize the amount of the poor-quality medications.
{"title":"Blockchain Applications to combat the global trade of falsified drugs","authors":"Y. Kostyuchenko, Qingshan Jiang","doi":"10.1109/ICDMW51313.2020.00127","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00127","url":null,"abstract":"The globalization of the pharmaceutical supply chain has lead to new challenges, the leading position among them is the fight against falsified and substandard pharmaceutical products. Such kind of products causes ineffective or harmful therapies all over the world. Traditional centralized technical tools can hardly satisfy the requirements of the changing industry. In this paper, we research the application of Blockchain solutions to modernize the drug supply chain and minimize the amount of the poor-quality medications.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121258094","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}
Advanced Metering Infrastructures (AMIs) facilitate individual load forecasting. The individual load forecasting not only improves the accuracy of aggregated load forecasting but is a fundamental component of various power applications. With the highlight of deep learning (DL) in the individual load forecasting, a serving platform specialized in deep learning is required to forecast with AMI stream data. However, the existing serving platforms for DL models do not consider stream data as an input but usually support image or text data through RESTful API. To solve this problem, we propose StreamDL that is a serving framework providing deep learning inference with AMI stream data. It leverages Apache Kafka to support stream data and Kubernetes to support the cloud environment. StreamDL considers the specific requirements for stream data, which supports stream parsing to fit any DL model especially recurrent network and continual training to alleviate accuracy degradation by the change of stream distribution. In this paper, we introduce the detail of the StreamDL platform and its use-cases using real AMI data.
{"title":"StreamDL: Deep Learning Serving Platform for AMI Stream Forecasting","authors":"Eunju Yang, Changha Lee, Ji-Hwan Kim, Tuan Manh Tao, Chan-Hyun Youn","doi":"10.1109/ICDMW51313.2020.00104","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00104","url":null,"abstract":"Advanced Metering Infrastructures (AMIs) facilitate individual load forecasting. The individual load forecasting not only improves the accuracy of aggregated load forecasting but is a fundamental component of various power applications. With the highlight of deep learning (DL) in the individual load forecasting, a serving platform specialized in deep learning is required to forecast with AMI stream data. However, the existing serving platforms for DL models do not consider stream data as an input but usually support image or text data through RESTful API. To solve this problem, we propose StreamDL that is a serving framework providing deep learning inference with AMI stream data. It leverages Apache Kafka to support stream data and Kubernetes to support the cloud environment. StreamDL considers the specific requirements for stream data, which supports stream parsing to fit any DL model especially recurrent network and continual training to alleviate accuracy degradation by the change of stream distribution. In this paper, we introduce the detail of the StreamDL platform and its use-cases using real AMI data.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127018210","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00046
J. V. D. Hoogen, Stefan Bloemheuvel, M. Atzmüller
Deep Learning (DL) provides considerable opportunities for increased efficiency and performance in fault diagnosis. The ability of DL methods for automatic feature extraction can reduce the need for time-intensive feature construction and prior knowledge on complex signal processing. In this paper, we propose two models that are built on the Wide-Kernel Deep Convolutional Neural Network (WDCNN) framework to improve performance of classifying fault conditions using multivariate time series data, also with respect to limited and/or noisy training data. In our experiments, we use the renowned benchmark dataset from the Case Western Reserve University (CWRU) bearing experiment [1] to assess our models' performance, and to investigate their usability towards large-scale applications by simulating noisy industrial environments. Here, the proposed models show an exceptionally good performance without any preprocessing or data augmentation and outperform traditional Machine Learning applications as well as state-of-the-art DL models considerably, even in such complex multi-class classification tasks. We show that both models are also able to adapt well to noisy input data, which makes them suitable for condition-based maintenance contexts. Furthermore, we investigate and demonstrate explainability and transparency of the models which is particularly important in large-scale industrial applications.
{"title":"An Improved Wide-Kernel CNN for Classifying Multivariate Signals in Fault Diagnosis","authors":"J. V. D. Hoogen, Stefan Bloemheuvel, M. Atzmüller","doi":"10.1109/ICDMW51313.2020.00046","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00046","url":null,"abstract":"Deep Learning (DL) provides considerable opportunities for increased efficiency and performance in fault diagnosis. The ability of DL methods for automatic feature extraction can reduce the need for time-intensive feature construction and prior knowledge on complex signal processing. In this paper, we propose two models that are built on the Wide-Kernel Deep Convolutional Neural Network (WDCNN) framework to improve performance of classifying fault conditions using multivariate time series data, also with respect to limited and/or noisy training data. In our experiments, we use the renowned benchmark dataset from the Case Western Reserve University (CWRU) bearing experiment [1] to assess our models' performance, and to investigate their usability towards large-scale applications by simulating noisy industrial environments. Here, the proposed models show an exceptionally good performance without any preprocessing or data augmentation and outperform traditional Machine Learning applications as well as state-of-the-art DL models considerably, even in such complex multi-class classification tasks. We show that both models are also able to adapt well to noisy input data, which makes them suitable for condition-based maintenance contexts. Furthermore, we investigate and demonstrate explainability and transparency of the models which is particularly important in large-scale industrial applications.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116010948","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00029
Venkataramana B. Kini, A. Manjunatha
This paper proposes and evaluates a multitask transfer learning approach to collectively optimize customer loyalty, retail revenue, and promotional revenue. Multitask neural network is employed to predict a customer's propensity to purchase within fine-grained categories. The network is then fine-tuned using transfer learning for a specific promotional campaign. Lastly, retail revenue and promotional revenue are jointly optimized conditioned on customer loyalty. Experiments are conducted using a large retail dataset that shows the efficacy of the proposed method compared to baselines used in the industry. A large retailer is currently adopting the proposed methodology in promotional campaigning owing to significant overall revenue and loyalty gains.
{"title":"Revenue Maximization using Multitask Learning for Promotion Recommendation","authors":"Venkataramana B. Kini, A. Manjunatha","doi":"10.1109/ICDMW51313.2020.00029","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00029","url":null,"abstract":"This paper proposes and evaluates a multitask transfer learning approach to collectively optimize customer loyalty, retail revenue, and promotional revenue. Multitask neural network is employed to predict a customer's propensity to purchase within fine-grained categories. The network is then fine-tuned using transfer learning for a specific promotional campaign. Lastly, retail revenue and promotional revenue are jointly optimized conditioned on customer loyalty. Experiments are conducted using a large retail dataset that shows the efficacy of the proposed method compared to baselines used in the industry. A large retailer is currently adopting the proposed methodology in promotional campaigning owing to significant overall revenue and loyalty gains.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128935401","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：