基于XGBoost和LSTM模型的高校招生网站流量预测研究  

Research on Prediction of Visits to College Admission Website Based on XGBoost and LSTM Model

作　　者：郑佳芳 游贵荣 ZHENG Jiafang;YOU Guirong(Information Technology Center,Fujian Business University,Fuzhou 350012,China)

机构地区：[1]福建商学院信息技术中心,福州350012

出　　处：《重庆科技学院学报（自然科学版）》2023年第5期45-50,共6页Journal of Chongqing University of Science and Technology：Natural Sciences Edition

基　　金：福建省中青年教师教育科研项目“基于XGBoost和LSTM模型的高校招生网站访问量预测研究”(JAT210384)。

摘　　要：为协助高校做好招生宣传工作,提出了基于机器学习的高校招生网站流量预测方法。首先,对网络日志进行预处理,生成不同时隙的数据集;接着,通过XGBoost模型的训练比较,筛选得出最佳实验数据集;然后,鉴于数据的非线性和趋势不一等特点,分别使用参数优化后的XGBoost和LSTM模型进行数据训练,并根据训练误差值计算权重系数;最后,应用XGB-LSTM加权组合预测模型进行数据预测。实验结果表明,该组合模型预测结果的平均误差分别比XGBoost和LSTM模型提高了80.28%和3.42%,具有良好的预测能力。To assist in the promotion of college admission,this paper proposes a method based on machine learning to predict visits to college admission website.Firstly,network logs are preprocessed to generate data sets with different time slots.Then,the best experimental data set is selected by comparing different XGBoost model training.After that,in view of the non-linearity and different trends of the data,the optimized XGBoost and LSTM models are used to train the data respectively,and the weight coefficients are calculated according to the training errors.Finally,XGB-LSTM weighted combined model is used for data prediction.The experimental results show that the average error of the combined model is 80.28%and 3.42%higher than that of the XGBoost and LSTM models respectively.It has good prediction ability and can provide data reference for college admission propaganda.

关 键 词：网站访问量 XGBoost LSTM 流量预测 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]