A new hybrid K-means evolving spiking neural network model based on differential evolution

Abstract

Clustering is one of the essential unsupervised learning techniques in Data Mining. In this paper, a new hybrid (K-DESNN) approach to combine differential evolution and K-means evolving spiking neural network model (K-means ESNN) for clustering problems has been proposed. The proposed model examines that ESNN improves by using K-DESNN model. This approach improves the flexibility of the ESNN algorithm in producing better solutions which is utilized to conquer the K-means disadvantages. Various UCI machine learning data sets have been utilized for evaluating the performance of this model. The results have shown that K-DESNN is much better than the original K-means ESNN in the number of pre-synaptic neurons measure and clustering accuracy performance.