Detection of Factors Affecting State Transition Based on Non-Homogeneous Markov Chain Model

Shen Xiujuan,Liu Hefei,Li Yong,Yuan Bin

doi:10.1109/access.2021.3097363

Abstract

The dependent and independent variables in traditional linear regression models are continuous numerical variables. When the dependent variable or independent variable is a discrete variable, the traditional linear regression model can no longer be used to analyze. To solve this problem, this article introduces the non-homogeneous Markov chain model. It introduces the mathematical definition of the non-homogeneous Markov chain model. And then this article uses Bayesian estimation method to derive posterior distribution of model parameters. Through the MCMC algorithm, we simulate an experiment, posterior means value of the parameters is estimated, and the estimation effect is found to be better. Finally, we analyze the impact of learning state transition about college students on the non-homogeneous Markov chain model. Influencing factors include whether to receive a scholarship and whether to serve as a class leader. In this paper, non-homogeneous Markov chain model is used to analyze and detect the impact of discrete variables on dependent variables. This is the major innovation in this article.

Highlights

The correlation and influence relationship between variables is important research content of statistics
The rest of this article is organized as follows: firstly, this paper introduces the mathematical definition of the non-homogeneous Markov chain model; secondly, it introduces the Bayesian inference method of the model; thirdly, it verifies the reliability of the estimation method through simulation experiments; it uses this method to analyze the influencing factors of College Students’ learning state transition
The linear regression model is the most classic statistical model to analyze the relationship between variables

Summary

INTRODUCTION

The correlation and influence relationship between variables is important research content of statistics. If the university students are class cadres or not, and whether the award-winning grants affect their learning status, we cannot directly model and analyze them by using the above-mentioned methods Under this background, this paper introduces and improves the hidden Markov model to analyze the interaction between discrete classification variables. On the basis of these studies, this paper improves the state transition mode of Markov chain and proposes a non-homogeneous Markov chain model, which is used to analyze the influencing factors of discrete classification variables. The rest of this article is organized as follows: firstly, this paper introduces the mathematical definition of the non-homogeneous Markov chain model; secondly, it introduces the Bayesian inference method of the model; thirdly, it verifies the reliability of the estimation method through simulation experiments; it uses this method to analyze the influencing factors of College Students’ learning state transition. We require setting prior distribution for each parameter and deriving posterior distribution

PRIOR DISTRIBUTION OF PARAMETERS

POSTERIOR DISTRIBUTION OF PARAMETERS

CONCLUSION