User Behavioral Preference Analysis Based on Location Information

According to the characteristics of users' location, the double-layer XGBoost algorithm is improved to infer the user's location information and realize the analysis of users' behavioral preference. Firstly, the outlier processing is proposed for the basic data such as shop information, user transaction records, etc.; secondly, the original shop location, average price and other information as well as the location information data such as GPS and WiFi of users' mobile devices are used to optimize the alternative set of precise location and build positive and negative sample. Finally, the improved XGBoost algorithm constructs the multi-dimensional features of multi information fusion to predict user behavior. The experimental results show that the double-layer XGBoost can achieve an average prediction accuracy of 92.65% on the user transaction record data set, which is better than SVM and KNN algorithm. Therefore, the double-layer XGBoost algorithm can achieve high accuracy and fast user behavioral preference analysis.