Markov Network Classification for Imbalanced Data with Adaptive Weighting

Abstract

This paper studies an imbalanced data classification algorithm based on a Markov network to solve the impact of severe imbalance in the distribution of class samples on classification performance. In view of the shortcomings of traditional methods in minority class identification, this paper introduces an adaptive class weight adjustment strategy and an optimized graph structure modeling method to improve the sensitivity and classification ability of the Markov network to minority class samples. The KDD Cup 1999 dataset is used in the experiment. The effectiveness of the proposed algorithm is verified by changing the proportion of minority class samples and testing data sets of different sizes. The results show that compared with traditional imbalanced classification methods, the algorithm has achieved significant improvements in accuracy, F1-Score, and AUC-ROC, especially showing strong robustness under extreme imbalance conditions. In addition, the computational complexity analysis shows that the algorithm has high efficiency in the inference stage and is suitable for large-scale data set applications. The research in this paper provides new ideas for solving the problem of imbalanced data classification and provides theoretical support and technical reference for the practice in related fields.