基于GBDT的美团菜谱推荐算法研究

Research on Meituan Recipe Recommendation Algorithm Based on GBDT

随着外卖行业的迅猛发展,外卖App对于菜品和商家的推荐服务精准度要求也水涨船高。因此,根据现有数据研究出了基于决策提升树的外卖推荐算法。在算法的设计研究过程中,通过对外卖订单数据的初步分析,根据其特点选取决策提升树作为算法的分类器。在具体模型方面,希望通过使用3种不同的模型,即XGBoost、CatBoost和LightGBM,来分别得到分类结果并分析研究在当前环境下性能最强的模型并给出相关建议。希望对于局部数据的研究能够在一定程度上代表长期的用户外卖订单数据,进而能够为推荐算法的研究提供一定的新思路。

With the rapid development of the takeout industry, the accuracy requirements of the takeout app for dishes and merchants’ recommendation services are also rising. Therefore, according to the existing data a takeaway recommendation algorithm based on decision lifting tree is developed. In the process of algorithm design and research, through the preliminary analysis of sales order data, the decision lifting tree is selected as the classifier of the algorithm according to its characteristics. In terms of specific models, we hope to use three different models, namely XGBoost, CatBoost and LightGBM, to get the classification results, analyze and study the models with the strongest performance in the current environment and give relevant suggestions. It is hoped that the research on local data can represent the long-term user take out order data to a certain extent, and then provide some new ideas for the research of recommendation algorithm.