摘要
餐饮企业营业额预测是一种根据相关特征参数构建机器学习模型并实现科学预测的人工智能技术,可以为企业管理层提供较为精准的营业额预测,使其合理控制经营成本,提高利润和综合竞争力.通过采集整理进驻安徽理工大学的高校餐饮企业历史经营数据,构建了一种优化随机森林算法的营业额预测模型,设计了一种二次划分贪心选择网格搜索算法,对随机森林算法中的子模型数量(n_estimators)和最大特征数(max_features)进行参数优化.实验结果表明,该模型在测试集上的预测均方根误差(RMSE)、平均绝对误差(MAE)和平方相关系数(R2)分别为0.037、0.149和0.963,经对比分析,明显优于传统的Ridge、SVM和XGBoost模型,证明提出的方法具有更好的预测效果.
Catering enterprises turnover forecasting is an artificial intelligence technology that builds a ma⁃chine learning model based on relevant feature parameters and implements scientific forecasting.It can provide enterprise management with more accurate turnover forecasting,reasonably control operating costs,and increase profits and overall competitiveness.By collecting and collating historical management data of catering enterprises in Anhui University of Science and Technology,a turnover forecast model based on optimized random forest algorithm is constructed,and a two-partition greedy selection grid search algorithm is designed,the number of sub-models(n_estimators)and the maximum number of fea⁃tures(max_features)are optimized for parameters.The experimental results show that the model's predict⁃ed root mean square error(RMSE),mean absolute error(MAE),and square correlation coefficient(R2)on the test set were 0.037,0.149,and 0.963,respectively.After comparative analysis,it is significantly better than the traditional Ridge,SVM and XGBoost models,which proves that the proposed method has better prediction effect.
作者
石文兵
苏树智
SHI Wen-bing;SU Shu-zhi(School of Computer Science and Engineering,Anhui University of Science and Technology,Huainan 232001,China)
出处
《通化师范学院学报》
2021年第2期88-94,共7页
Journal of Tonghua Normal University
基金
国家自然科学基金项目(61806006)
安徽省高等学校自然科学研究项目(KJ2018A0083)。
关键词
随机森林算法
营业额预测
高校餐饮
参数优化
网格搜索
random forest algorithm
turnover forecast
college catering
parameter optimization
grid search
