Research on the College Student Employment Prediction Model Based on Random Forest and Deep Belief Network

To help schools and students better respond to the dynamic job market and improve the alignment between education and employment, this paper proposes a hybrid model for student employment prediction that combines an improved Random Forest (RF) algorithm and a Deep Belief Network (DBN). First, student data is preprocessed to remove irrelevant attributes to achieve data consistency. Second, to enhance the accuracy of the prediction model, a feature selection model that integrates Principal Component Analysis (PCA) and the RF algorithm is employed to select the optimal subset from the original features. Finally, the selected features are used as input for the DBN to learn high-level feature representations and complete the student employment prediction. Experimental results indicate that the proposed hybrid model achieves a prediction accuracy of over 91 %, which is 15.6 % and 13.9 % higher than that of Recurrent Neural Network (RNN) and Multi-Layer Perceptron (MLP) models, respectively. This model can assist educational institutions in understanding employment trends across different majors and courses, thereby improving students' career development path planning.