Assistant Decision-making System In Conversion Network Based On Feature Engineering And Multi-model Detection

In recent years,under the country focuses on the macroeconomic situation and international development trends,Ministry of Industry and Information Technology issued documents that required that the user numbers of domestic operators can be freely converted without changing the network type,that is Conversion network with number.Under the new business competition model,how to give full play to the value of massive data in the communications industry,and retain more local customers and attract more foreign customers,it is key to apply artificial intelligence and big data technology to predict customer transfer behavior and assist operators in making business competition decisions or building machine learning algorithms that portability is predicted to succeed.However,among the current risk prediction models,for the communications industry,there are many shortcomings and shortcomings in practical applications.The main contents of this thesis are as follows:1.In order to improve the prediction effect and enrich the information of data,data preprocessing and multi-feature construction are carried out.In this thesis,data partition is constructed after missing value,time and information redundancy and outlier handling processing.Construct a large number of features,and perform feature selection and feature classification according to the actual significance of feature importance.2.To further improve the forecasting effect,parameter tuning and multi-model building and multi-model fusion strategy are carried out.Due to insufficient upper limit of single model generalization ability,we used Multi-model fusion takes advantage of each model to build multiple single-model stacking of GBDT,Rf,Extremely randomized tree,Adaboost and build XGBoost as the second-layer single model after fusion,for raising the upper limit of overall generalization ability.3.Taking the above feature engineering and parameter tuning,multi-model fusion and multi-feature construction as the core to design and implement the assistant decision-Making System in conversion network,assists managers of major operators in making business competition decisions.The dataset of this thesis is derived from Chongqing Mobile users' real behavior data from December 2019 to February 2020 and it was anonymized,and the evaluation standard is the F1-Score(It is the harmonic mean of precision and recall.The maximum is1 and the minimum is 0.The more accurate the model,the larger the F1-Score value).In this thesis,F1-Score is improved to 0.755 through data preprocessing,feature construction and feature selection and parameter tuning,and through multi-model fusion,the prediction effect is significantly improved.The F1-Score is improved to 0.7607.This article is based on the above core methods,designed and implemented the assistant decision-Making System in conversion network,assisting operators in monitoring the status of number portability and network transfer in all aspects,enhancing business competitiveness and precise user experience.