Prediction of marketing strategies performance based on clickstream data

Today, Internet-based businesses are one of the most useful tools to make gain in the economies of developing and developed countries. It can even said that the expansion of the World Wide Web caused other businesses to seek customers in the virtual advertising and online world to increase their sales. This study presents a data-driven approach to predict the success of the marketing strategies performance of an online shopping store. The data has been collected by a Poland online shopping website in the year 2008, which has extracted in the UCI datasets. In the data preparation phase, a decision tree (DT) is developed and 13 features of customers are selected for modeling phase. In the proposed method in this research, the rminer package of R software is used. In which three classification models including neural network(NN), support vector machine (SVM), and logistic regression(LR) are developed. Then, two criteria of AUC and ROC curves are used to compare these three models. By comparing the models, it is determined that the NN technique works better than the other three models in prediction. This result can be helpful for marketing managers to plan effectively in website design to attract new visitors and shoppers.