Hybrid Knowledge Extraction Framework Using Modified Adaptive Genetic Algorithm and BPNN

Fault diagnosis based on the expert system (ES) is still a research topic of manufacturing in Industry 4.0 because stronger interpretability. As core component ES, fault accuracy positively correlated to precise knowledge base. But it difficult for users understand obtained from original dataset utilizing existing extraction method. Therefore, great significance extract easy-to-understand and exact rules NN framework. This paper proposes hybrid framework perform rule overcoming this drawback. First, an improved adaptive genetic algorithm (GA) using logistic function, namely LAGA, proposed solve traditional GA’s insufficient prediction performance issue. Compared with other three mainstream GAs, experiment results optimizing six selected test functions by these GA variants show that LAGA algorithm’s convergence speed have been greatly improved, especially high latitude functions. On basis, method symbol NN, LAGA-BP framework, discussed manuscript classify real-valued attributes. obtains hidden (knowledge refinement process) further transforms acquired into more (rule process). The execution could be separated two phases. first phase optimize back propagation (BPNN) refine classification over optimized BPNN. In second phase, attribute reduction multi-layered (SD algorithm) different superposed networks used reduce data-set attributes then uses K-means clustering if-then simplified Wisconsin breast cancer as case study reveal correctness robustness Consulting relevant medical personnel referencing data shows extracted help verify results, thus verifying framework’s feasibility practicality.