Research on Logistics Cost Prediction Based on Random Forest Regression Modeling

Abstract

With the rapid development of globalization and e-commerce, logistics industry has become a key link between production and consumption. However, the efficient management and control of logistics costs has become an important challenge for enterprises to enhance their competitiveness. The scientific control of logistics cost depends on accurate cost prediction. In this paper, we first process the missing values of open source logistics cost data, and analyze the logistics cost and numerical influencing factors as skewed distribution with the help of distribution probability density plot and histogram, so we use logarithmic transformation to eliminate the skewness. The box plot and quartile method are used to visualize and process the outliers of discrete factors. For the cleaned data, the chi-square test and correlation coefficient method are applied to screen important features and remove redundant information, respectively. Then the random forest regression model was constructed to predict logistics costs, and the model parameters were optimized by grid search and cross-validation. Finally, the importance of the features of the prediction model is also ranked, and the model prediction mechanism is analyzed in depth to provide a scientific basis for logistics cost control.