Recurrent Neural Networks based Clustering for Binary Data in S-P Charts
Kazuma Kiyohara, Toshimichi Saito · 2024
This paper presents a novel application of recurrent neural networks to clustering for binary data in S-P charts. The S-P chart is a simple/useful binary data set in instruction of students. The chart consists of binary score vectors of students and is characterized by several feature quantities. However, as the number of students increases, the S-P chart becomes complicated and the instruction becomes harder. In order to divide the large S-P chart into suitable small S-P charts for instruction, we present a clustering method of the binary score vectors. The method is based on nonlinear dynamics of the recurrent neural networks having multiple stable fixed points. Basin of attraction to one fixed point corresponds to one cluster. Performing numerical experiments for typical example of artificial large S-P charts, effectiveness of the proposed method is confirmed.