Probabilistic-based Markov chains for behavioral prediction

Author:
Open Access
Publisher website
Google Scholar
Abstract
Due to the wide application of Markov chains, it makes some models that cannot be computed due to a large amount of computation have an approximation. In this paper, based on Markov, combining probability theory with a state transfer probability matrix and using the ordered clustering method to divide the behavior into clusters, we construct a behavioral prediction model based on the probabilistic Markov chain to solve the problems that the model tends to have such problems as low overall prediction accuracy and limited applicability. By testing the model’s performance on the relevant dataset, we can predict the occupants’ in-room status. The Gowalla dataset has an MMP model that is 16% accurate and 21% recall. Classifying households and identifying indoor behavior patterns of different households is sufficient so that the indoor behavior patterns of the same type of households are closer to each other. The method is capable of considering various household characteristics parameters and their influence on in-room behavior comprehensively and classifying actual behavior reasonably.

This publication has 14 references indexed in Scilit: