Extended belief rule-based system (EBRBS) as an integrated data and knowledge driven rule-based system has attracted much attention in the last few years and has been widely used in classification problems. Data inconsistency and data incompleteness are two common issues and result in the decrease of the accuracy of data-driven model including EBRBS. Although a dynamic rule activation (DRA) method was proposed to solve these two issues by selecting the most consistent rules and has shown its capability in improving the performance of EBRBS, there still exists some drawbacks in its efficiency and rationality. Hence, a new DRA method based on activation factor (called AFDRA) is proposed for EBRBS to better handle the data inconsistency and data incompleteness issues. Case studies show that the AFDRA method not only has a great improvement in the term of efficiency over the DRA method, but also achieves better performance for EBRBS in classification problems.
|Title of host publication||2021 16th International Conference on Intelligent Systems and Knowledge Engineering (ISKE)|
|Number of pages||5|
|ISBN (Electronic)||978-1-6654-0553-9, 978-1-6654-0552-2|
|Publication status||Published - 18 Apr 2022|
|Event||16th International Conference on Intelligent Systems and Knowledge Engineering - Chengdu, China|
Duration: 26 Nov 2021 → 28 Nov 2021
|Name||2021 IEEE International Conference on Intelligent Systems and Knowledge Engineering, ISKE 2021|
|Conference||16th International Conference on Intelligent Systems and Knowledge Engineering|
|Period||26/11/21 → 28/11/21|
Bibliographical noteFunding Information:
This research was supported by the National Natural Science Foundation of China (Nos. 72001043 and 71872047), the Natural Science Foundation of Fujian Province of China (No. 2020J05122), the Humanities and Social Science Foundation of the Ministry of Education of China (Nos. 20YJC630188 and 19YJC630022), the Social Science Planning Fund Project of Fujian Province of China (No. FJ2019C032).
© 2021 IEEE.
- activation factor
- dynamic rule activation
- extended belief rule based system