Research and Prediction of Influential Factors of Film Box Office: Based on Machine Learning Algorithms Such as XGB, LGB and CAT
Boning Jiang · Frontiers in artificial intelligence and applications · 2024
In response to the issue of film box office predictions, in this paper, we first built XGBoost, and used the training set to optimise the model by grid search and cross-validation for many times of tuning.Then we obtained the model prediction results. After that, we built the LightGBM model, tried to add text features, variable combinations, and interactions in turn, introduced the SHAP model to judge the specific impact of each important feature, and used the tuned model for training and comparison to evaluate its performance on the test set. In addition, we built the CatBoost model following the same method. According to the rmse comparison it was found that lightGBM had the best prediction. Finally, the models were fused by the Blending method, and the fused models were found to have better prediction results.