Application Of XGBoost Algorithm In Sales Forecast Of Smart Home Appliances

As the growth of cloud computing and big data,intelligent manufacturing will become the mainstream of future manufacturing development.More and more traditional manufacturing films pay attention to the interconnection and data analysis between equipment,and promote the intelligent development of the company in all aspects by combining their traditional information technology with the Internet and big data,so that the company is in a leading position in the market.In the field of traditional home appliances,it is not only necessary to make home appliances intelligent,but also to continuously inject intelligent knowledge into the production and sales process of home appliances.In this way,the manufacturing and sales process can become intelligent,and the performance of home appliances can also be improved.As smart home appliances become more and more popular in the market,many household appliance enterprises have prodeced massive amounts of data.It is possible to tap the potential value of home appliance data,and stimulate the market potential of the home appliance industry,by storing,classifying,predicting,and analyzing massive amounts of home appliance data through the big data platform.The paper’s main contents are listed as follows:(1)The paper’s research topic is the sales data and store information data of Midea Entrepreneur Electric Store.Facing the growing data of smart home appliances,the Hadoop is used to store a large amount of home appliance data.Due to the weak iteration of the Map Reduce computing framework under the Hadoop platform,the advantages of the Hadoop platform and the Spark computing framework are fully utilized,thereby improving the efficiency of data processing,by using the Spark computing framework on the Hadoop platform to perform distributed computing on the data.(2)On the grounds of the data of the data information of smart home appliances,it is possible to visualize the data,analyze the impact of related variables on the sales of smart home appliances,sort the data features on the importance,and establish a good feature equation.At the same time,the paper introduces linear regression,Light GBM,XGBoost and other model algorithms.The author compares and analyzes those algorithms in theory.A single XGBoost model has great advantages in processing smart home appliance data.(3)In order to improve the accuracy and generalization ability of the XGBoost model for predicting sales,the Stacking theory of algorithm is used for optimizing the XGBoost model,the XGBoost and Light GBM models are integrated,and features with high importance is added for training in the feature selection of the second layer of Stacking.The experimental results show that the prediction performance of the optimized XGBoost model is further improved.The accuracy and generalization ability of the optimized XGBoost model are better,and the prediction results can be better fed back to the production department of the enterprise in comparison with the single model,so that the inventory of related products can be handled well in advance.