Research on the rating model of poor vocational college students based on ADA-MDF algorithm

Given the significant workload involved in poverty assistance at domestic colleges and universities, as well as the challenges posed by subjective factors, this paper proposes an ADA-MDF model for the prediction of poverty ratings among students. Initially, the cascade structure of deep forest is extended, and an MDF model is developed by selecting different combinations of five base learners. Subsequently, using an over-sampling algorithm, the SMO-MDF and ADA-MDF models are constructed. A comparative analysis is performed with the baseline MDF model (without sampling), ultimately demonstrating that the ADA-MDF model achieves superior overall performance. Detailed analysis of the rating results is also conducted. The experimental results show that the ADA-MDF model significantly improves the performance of all evaluation indexes. In addition, the in-depth analysis of the model and data, as well as the comparison of the feature selection results, emphasize the importance of feature selection in improving the prediction accuracy of the model.