Application of Hysteresis Neural Networks to Clustering for S-P charts

This paper studies application of the hysteresis neural network to clustering for the S-P charts. The network is characterized by binary hysteresis activation function and the dynamics is described by a piecewise linear differential equation. The stable equilibrium points are calculated exactly. The S-P chart is a binary data set consisting of score vectors of students. The chart provides useful information to help the education. However, as number of students increases, the S-P chart becomes hard to handle. We present a clustering method that classifies the large-scale S-P chart into small-scale S-P charts. The basin of attraction to equilibrium points corresponds the clusters. Executing fundamental numerical experiments, the clustering performance is confirmed.