Research and Application of Prediction Model Based on Random Forest Algorithm
Jing Geng, Yifan Zhai · 2023
In this paper, based on the open source UC dataset, the original dataset was first subjected to data preprocessing operations such as logarithmic transformation, outlier and missing value processing, and dimensionality reduction to obtain complete, valid, intuitive, and concise data. To avoid multicollinearity among variables, the variance expansion coefficient of each variable was calculated and found to be less than 5, so there was no multicollinearity. Then lasso regression feature extraction was performed to reduce the number of variables to 13, thus reducing the complexity of the model. A random forest prediction model was constructed, and its average relative error was 0.04217 according to the model evaluation criterion 0.99157, with an error precision of 1. The model has good precision and is useful for prediction research in other fields.