Enze Wu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Enze Wu has a total of 6 co-authors. Here are the top co-authors by number of shared publications:
- Peiyi Tang
Nanjing Huihai Transportation Technology Company Limited, China
1 shared publication - Taizhi Lv
School of Information Technology, Jiangsu Maritime Institute, Nanjing 211170, China
1 shared publication - Tao Huang
School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing, China
1 shared publication - Yingying Chen
School of Information Engineering, Jiangsu Maritime Institute, Nanjing, China
1 shared publication - Zhidan Yuan
College of Information Engineering, Jiangsu Maritime Institute, Nanjing, Jiangsu, China
1 shared publication - Zhuangzhuang Wang
College of Information Engineering, Jiangsu Maritime Institute, Nanjing, Jiangsu, China
1 shared publication
Latest Publications
- Open Access
- 11 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 43-52
Empirical Studies on Failure Prediction for Distributed Systems Based on Feature Selection
Distributed system failure prediction identifies potential failures by constructing machine learning models based on Key Performance Indicator (KPI) data. However, due to the diversity of KPI metrics, not all metrics are relevant to specific failures. This lead to redundancy in the KPI metrics used to construct the models. To address this issue, we investigate the impact of filter-based ranking feature selection methods on distributed system failure prediction. In our empirical study, we conduct experiments based on the ZTE dataset by using six feature selection methods and four tree-based models. We explore the effects of different feature selection methods and selection ratios on model performance. Additionally, we study whether feature selection methods can effectively identify important KPI metrics from the perspective of model interpretability. The results indicate that compared to not using feature selection methods, employing such methods allows for the construction of effective failure prediction models with only a subset of KPI metrics. Model performance improves with an increase in feature selection ratio until reaching a point of stabilization. Moreover, feature selection methods effectively identify important KPI metrics within the dataset.
- 11 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 43-52
Open AccessEmpirical Studies on Failure Prediction for Distributed Systems Based on Feature Selection
Distributed system failure prediction identifies potential failures by constructing machine learning models based on Key Performance Indicator (KPI) data. However, due to the diversity of KPI metrics, not all metrics are relevant to specific failures. This lead to redundancy in the KPI metrics used to construct the models. To address this issue, we investigate the impact of filter-based ranking feature selection methods on distributed system failure prediction. In our empirical study, we conduct experiments based on the ZTE dataset by using six feature selection methods and four tree-based models. We explore the effects of different feature selection methods and selection ratios on model performance. Additionally, we study whether feature selection methods can effectively identify important KPI metrics from the perspective of model interpretability. The results indicate that compared to not using feature selection methods, employing such methods allows for the construction of effective failure prediction models with only a subset of KPI metrics. Model performance improves with an increase in feature selection ratio until reaching a point of stabilization. Moreover, feature selection methods effectively identify important KPI metrics within the dataset.
- Open Access
- 13 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 840-844
Research on real-time processing of vessel transportation risk based on stream computing
Vessel transportation safety involves various factors such as shipping plan, vessel certificate, vessel identification, automatic identification system (AIS), vessel pollutant discharge, vessel shore power and so on. The identification and supervision of vessels entering a jurisdiction will help identify safety risks in transportation and take corresponding measures to ensure shipping safety and the ecological environment. In order to deal with the vessel safety in real-time, a real-time data processing system for vessel transportation risks based on stream computing is designed and implemented. This system uses the Flink platform to process vessel risk information in real-time, and the processed risk information is stored in Redis to provide efficient query. Historical vessel risk information is stored in Doris, providing fast history search. The system has been applied in a certain section of the Yangtze River, and the usage results prove that the it can effectively deal with vessel transportation risks in real-time and improve the intelligent level of maritime supervision.
- 13 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 840-844
Open AccessResearch on real-time processing of vessel transportation risk based on stream computing
Vessel transportation safety involves various factors such as shipping plan, vessel certificate, vessel identification, automatic identification system (AIS), vessel pollutant discharge, vessel shore power and so on. The identification and supervision of vessels entering a jurisdiction will help identify safety risks in transportation and take corresponding measures to ensure shipping safety and the ecological environment. In order to deal with the vessel safety in real-time, a real-time data processing system for vessel transportation risks based on stream computing is designed and implemented. This system uses the Flink platform to process vessel risk information in real-time, and the processed risk information is stored in Redis to provide efficient query. Historical vessel risk information is stored in Doris, providing fast history search. The system has been applied in a certain section of the Yangtze River, and the usage results prove that the it can effectively deal with vessel transportation risks in real-time and improve the intelligent level of maritime supervision.