An algorithm derived from the mixture probability algorithm has been constructed by using the autoregressive model(AR) with bias component to analyze the cross-sectional mean void fraction signals. However in the algorithm with a maximum likelihood estimates, equations for parameter estimation were very complicated. So in this paper, we modified the mixture probability algorithm and proposed new algorithm based on a stochastic Newton method.The effectiveness of our method is confirmed by applying ones to classify the crosssectional mean void fraction signals of gas-liquid two-phase flow.