Deep belief networks in understanding and predicting changes in students’ political attitudes

Abstract

Abstract Leveraging data collection software, this study gathered and preprocessed text data reflecting students’ political attitudes across five dimensions: political cognition, ruling party cognition, positive political emotion, negative political emotion, and political behavioral tendency. The processed data were subsequently stored in Excel format. These dimensions of change in students’ political attitudes served as inputs for a deep belief network (DBN) model, with the changes in attitudes being the outputs. The model’s parameters were meticulously tuned to optimize its predictive capabilities, aimed at effectively constructing a prediction model for shifts in students’ political attitudes using the deep belief network. To refine the model’s parameters and minimize errors during inverse fine-tuning, the chimpanzee algorithm (ChOA) was employed. This novel approach, termed ChOA-DBN, facilitated the successful construction of the predictive model for students’ political attitude changes. The model’s efficacy was assessed using the collected data on students’ political attitudes. Notably, the DBN model optimized via the chimpanzee algorithm demonstrated robust predictive performance, particularly when processing sample points exhibiting significant changes in students’ political attitudes (ranging from 20 to 45). The model’s predictions predominantly fell within a ±15% error margin. This research not only showcases the accuracy of the ChOA-DBN model in predicting alterations in students’ political attitudes but also underscores its potential contribution to fostering the development of accurate political perspectives among students.