Design And Implementation Of Churn Prediction System For Operators Based On Cloud Computing

Nowadays with computer technology’s development, the data we own is increasing at an exponential growth. While single CPU node is becoming weaker for processing big data, using cloud computing with clusters for data storage and data mining is a new trend. Carriers are common owners of big data who hold resources of customers’ information and behavior. As the technology matures, monopoly is broken and the competition between operators is becoming fiercer. Customers are free to choose cheaper expense and better service which always causes churn. Under this circumstance, the thesis designs and implements the churn prediction system in a cloud computing environment.Based on the function requirement of the system, this thesis firstly designs the overall framework of the prediction system. Secondly it gives the design of data processing module, model training module, model testing module and model applying module. Thirdly it implements each module. At last it tests and analyzes the performance of the system.In this thesis, the prediction system is using Logistic Regression, Na?ve Bayes Classifier and BP Neural Network as theoretical basis. And it is focusing on the parallelism of three algorithms in Hadoop environment. It uses training dataset to obtain prediction models, then uses testing dataset to test the results. It is found that all the three algorithms have their advantages and disadvantages in different scenarios. BP Neural Network works best particularly for telecom operators, so the prediction system applies BP Neural Network to practical use. And the results of the system are given at last.