Metrics Prediction and State Transfer Analysis Based on XGBoost and Markov Chains

View at publisher Summarise with VIRAT
Authors: Conghao Liang, Zitai Zhang
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In this paper, a multifactor feature model and a joint gray-Markov model are proposed, focusing on the variable influence mechanism and state transfer probability. Firstly, the multifactor model containing the calculation of award probability, strength index and total expected value is constructed, the XGBoost algorithm is introduced to process the feature vectors, the results are outputted through regression or classification strategy, and the significant positive correlation between variables and total indicators is verified by Spearman correlation analysis and OLS regression, in which the influence of the variables of the quantity category is stronger than that of the type category. Second, for the zero-indicator state group, the time series data are cumulatively generated, differential equations are solved and inverse operations are reduced by the gray model to construct the state space and define the two types of states. Finally, the state transfer probability matrix is computed using Markov chains, and the initial distribution is combined to predict the future state distribution. The model can quantify the effects of multidimensional variables, improve the fitting accuracy through parameter training and error control, and realize the effective modeling of nonlinear relationships in small samples and non-complete information scenarios, so as to provide methodological support for the prediction of indexes and the analysis of state transfer in dynamic systems.

Fields of Study & Subjects

Grey System Theory Applications Forecasting Techniques and Applications Data Mining and Machine Learning Applications

Publication details

Journal / source
IEEE International Conference on Power, Electronics and Green Energy (ICPEGE)
Pages
1648–1652
Publication date
28 January 2026
Language
EN