Research And Application Of Algorithm Based On Combination Of AdaBoostSVM

With the rapid development of information technology,the supermarket retail industry is under tremendous pressure from the e-commerce industry.Nowadays,supermarkets almost set up their own management system of goods entry,storage and sale.It has accumulated huge customer consumption data.There is a lot of valuable information hidden in these huge consumption data.It assists supermarket managers to make right business decisions by deeply analyzing consumption data.But it costs lots of effort and time to extract potentially useful information in huge information by means of artificial.Therefore,it already has become a hot topic of research to apply some algorithms in field of artificial intelligence to dig valuable information from huge records about consumption.In order to use mass consumption data,we design a model forecasted potential customers to predict customers with shopping potential in the paper.First of all,it acquires features affecting shopping potential by analysing customers' sales data,then selects training samples according to geometric distance between sample points and optimum separation hyperplane.Secondly,an adaboost algorithm improved way of updating weight is applied to combine SVM weak classifier to form an AdaBoostSVM combination algorithm.Follow,it proposes an AdaBoostSVM model based on selected samples and combination algorithm,which realizes to mine customers with shopping potential according to their records of consumption.Finally,real consumption data of certain supermarket were used to test the model.Experimental results show that compared to traditional classification algorithms,the proposed model in this paper has advantages of higher classification accuracy and less training time.Finally,we apply the proposed model based on Hadoop platform to predict customers with shopping potential in supermarket entry,storage and sale.It assists supermarket managers to make targeted marketing strategies.