Research on a Multi-dimensional Evaluation Model Based on an Intelligent Scoring Algorithm Using Entropy Weighting

Abstract

This paper focuses on the research of intelligent scoring algorithms, constructing a multi-dimensional evaluation model and intelligent scoring scheme based on the entropy weight method. The study first analyses the distribution characteristics of evaluation data from human reviewers and two AI algorithms, establishes a statistical model for evaluation data, and reveals the patterns of differences in scoring consistency across different question types. Subsequently, an evaluation indicator system is constructed based on accuracy, stability, and adaptability. Information entropy is used to quantify indicator variability and weights, establishing a comprehensive evaluation model that confirms the performance degradation of AI algorithms in subjective question scenarios. The study was then expanded to six academic disciplines, constructing a discipline-weighted scoring model to further analyse and compare the scoring effectiveness of the two AI algorithms across different disciplines. Finally, an intelligent scoring scheme was designed by combining error thresholds with dynamic weights based on the entropy weight method, validating the model's effectiveness across multiple disciplines and providing a data-driven objective evaluation framework for intelligent educational scoring.