Design and Optimization of Modular Neural Networks based on Cluster Analysis: Applied to Complex Data Pattern Recognition

The recognition of complex data patterns in a typical application is always a difficult topic. From the perspective of designing modular RNN and optimization based on cluster analysis to improve accuracy and stability in complex data pattern recognition, many researchers have taken actions in this direction. This article proposes a module RNN based on the method of cluster analysis in order to classify and identify patterns in complex data. Three methods, namely RNN, CNN, and LSTM, are employed in the experiment to evaluate three aspects of algorithm accuracy, recognition time, and algorithm stability. Complex data pattern recognition has brought significant improvement in the modular RNN based on cluster analysis. Secondly, from the perspective of accuracy, this method has good application value: experimental accuracy ranges from 92% to 98%. Lastly, the algorithm stability ranges from 94% to 100%, realiably showing that the overall performance of the proposed method is consistently strong on a range of different tasks and datasets.