Personalized recommendation model of traffic package based on user consumption behavior

Under the dual role of technological innovation and market competition, the telecom industry is facing extremely severe challenges. The domestic telecom market is gradually saturated, and the competition among telecom operators is becoming increasingly fierce. The telecom package service is the core business for operators to acquire customers and enhance their value. How to effectively use the value of the package, reduce the customer churn rate and promote the value-added service, and recommend suitable telecommunication packages for users, which is of great significance to telecom operators. This article completes the telecom package recommendation in four phases. First, data processing of real user traffic package data from China Unicom Research Institute, extracting information from the actual consumption data and consumption characteristics of a large number of existing users, and dynamically analyzing user needs; secondly, model building, using support vectors Machine (SVM), Logistic Regression, Decision Tree Model as the baseline, and then choose two tree-based algorithms that are used in practical applications for comparison. One is XGBoost [1], and the other is LightGBM [2], proposed telecom package personality recommendation based on these two models; third, model training, using the existing training set to train our built model, and then using the test set to test, the predicted results and actual results The user traffic plan is compared and the accuracy is calculated. Fourthly, the optimization result, the data is further processed, and the model is improved. It is found that the accuracy of the stacking fusion model is better than that of the single model. The experimental results show that our model can accurately recommend the appropriate traffic package to users.