基于Bayesian-Stacking模型的电影票房预测
Prediction of movie box office based on the Bayesian-Stacking model
李小红 1韩淑淑2
作者信息
- 1. 山东工商学院数学与信息科学学院,山东烟台 264005
- 2. 山东工商学院统计学院,山东烟台 264005
- 折叠
摘要
本文构建了一种基于XGBoost的特征选取方法以及Bayesian-Stacking集成算法的票房预测模型.首先,构建 XGBoost 的影响力测量模型进行变量筛选,能够简化后期模型的输入和提高模型特征变量的可解释性;其次,分别构建了BP神经网络、XGBoost、Logistic Regression、LightGBM、GBDT以及Stacking模型,再利用贝叶斯优化算法实现上述模型超参数全局寻优后,对电影票房进行预测;最后,引入评价指标进行分析.结果表明:1)将贝叶斯优化算法与模型相结合,获得了相对于原模型更高的预测精度;2)Bayesian-Stacking模型的电影票房预测精度均优于其他模型.Bayesian-Stacking模型在电影上映期间预测最终票房具有较高的参考价值,可为有关部门提供决策参考.
Abstract
This paper constructs a feature selection method based on XGBoost and a box office prediction model based on the Bayesian-Stacking integrated algorithm.Firstly,constructing XGBoost's influence measurement model to screen variables can simplify the input of the later model and improve the interpretability of the model's characteristic variables;Secondly,BP neural network,XGBoost,Logistic Regression,LightGBM,GBDT,and Stacking models are constructed respectively,and then the box office of the film during the release period is predicted after the global optimization of the above models is realized by Bayesian optimization algorithm.Finally,the evaluation index is introduced for analysis.The results show that:1)the Bayesian optimization algorithm is combined with the model,and higher prediction accuracy is obtained compared with the original model;2)the Bayesian-Stacking model is superior to other models in box office prediction accuracy.The Bayesian-Stacking model has a high reference value in predicting the final box office during the film release period and can provide decision-making reference for relevant departments.
关键词
应用统计数学/电影票房预测/Stacking模型/XGBoost/贝叶斯算法Key words
applied statistical mathematics/box office forecast/Stacking model/XGBoost/Bayesian algorithm引用本文复制引用
基金项目
山东省自然科学基金(ZR2021MA007)
出版年
2024
NETL
NSTL
万方数据