Analysis And Forecast Of Mobile Phone Sales Based On Data Mining

Under the background of the rapid development of the Internet,offline payment,e-commerce,O2 O,social media and other industries are constantly rising.Smart phones play a vital role in it,and more and more influence and change people's lives and work,and gradually become a necessary mass consumer product.Due to the rapid development of e-commerce industry,people's shopping mode has also undergone significant changes,from the traditional offline purchase to today's online purchase,opening up a new sales channel for the mobile phone market.From the point of view of e-commerce,the problem of mobile phone sales has a great impact on the revenue of operation,the determination of commodity inventory and the formulation of marketing plans.If the sales volume of various types of mobile phones can be better predicted by establishing models,e-commerce can determine reasonable inventory to maximize revenue,and timely access to user's needs,grasp the future market trends of goods,and also facilitate customers to select and purchase various types of goods.In recent years,data mining algorithms and machine learning technology have been maturing and gradually applied widely.Decision tree,random forest,neural network,support vector machine(SVM)and other methods are different from the traditional linear regression model.They do not need to satisfy the assumption of data distribution,and the prediction effect is ideal,so they are gradually applied in statistical prediction model.Because there are many factors to be considered in the prediction model of mobile phone sales,and the model may be more complex,so this paper will use data mining method to model and analyze.This thesis collects 3327 mobile phone information from an e-commerce website based on network crawling method,and selects 32 key factors affecting mobile phone sales,including sales volume,price,brand,memory,operating system and so on.Firstly,the data is preprocessed,and the Boruta algorithm and Lasso method are used to screen the variables preliminarily,eliminating the factors that contribute little to the model,so as to reduce the complexity of the model,and finally 21 variables are selected for further analysis.Then the prediction model of training set data is established.The main models are support vector regression model,BP neural network regression model and random forest regression model.After adjusting the parameters repeatedly,the best fitting effect of each model is achieved.In order to evaluate the performance of the model,a fivefold cross-validation method is used to compare the prediction results of the three models.The results show that the random forest regression model has the best effect and the best performance.Finally,the test set data is used to test the prediction errors of the three regression models.By comparing the accuracy evaluation indexes,it shows that the random forest regression model is the best one,which has a certain reference value for mobile phone sales prediction.