An efficient algorithm for clustering data and best model selection using AIC

This paper introduces Binary Synaptic Weights version 2 (BSW2), an efficient neural network algorithm that can cluster a large number of data to substantially large numbers of bins. The weight data of BSW are represented in binary and that, most importantly, it guarantees to learn. In this paper BSW is further developed by introducing a few number of parameters which affect the number of bins to be created as well as the performance of the model in terms of the accuracy of the model´s prediction. Thus, a number of models are created according to certain combinations of these parameters for the same data set. In order to select the best model among those models, the AIC (Akaike information criterion) is employed but modified by adding a new term corresponding explicitly to the accuracy of each model´s prediction. Actual stock market data are used to test these models´ performances and a practical method of selecting the best model is presented