Research on integrating different methods of neural networks with case-based reasoning and rule-based system to infer causes of notebook computer breakdown

Important issues for notebook computer companies include how to ascertain the problems of machines sent by customers, and then assigning those machines to the appropriate department for servicing; and how to maintain breakdown data to save both handling time and costs. However, in practical application, unreliable data decreases the model’s accuracy, and thus, new methods are brought forward in rapid succession to increase accuracy when inferring causes of notebook computer breakdown. This study integrated several different methods, consisting of a neural network, with case-based reasoning (CBR) and a rule-based system (RBS) to propose a gradual model for inferring causes of notebook computer breakdown. It stressed that the model should have accuracy, elasticity, and transparent interpretability. The model contains three phases: data extracting, group indexing and knowledge creation. Initially, the data extraction phase uses a self-organizing map (SOM) and a revised learning vector quantization network method to reduce isomorphic data to similarity characteristic-based clustering, thus, improving data quality. Then, the group indexing phase establishes a clustering index prediction model based on a back-propagation network (BPN) and genetic algorithm (GA) to increase the efficiency of case selections. Then, the knowledge creation phase uses CBR and RBS to create a notebook computer breakdown case selection model to determine the breakdown cause. Finally, the experimental results show that data purification can actually improve the model’s accuracy. The CBR with clustering index and rule-based reasoning has a better classification accuracy rate than either the CBR, without the clustering index and rule-based reasoning, or the traditional CBR, in addition, it provides a reference for inferring causes of notebook computer breakdown.