AISWare AI² helps a provincial operator to build machine learning platform

Customer requirement

As the market environment changes, a provincial operator is facing challenges such as user churn, insufficient traffic vitality, weak revenue growth, and poor customer perception. Conventional empirical analysis and operation methods have been unable to realize the precise screening of target customers, the precise matching of product strategies, and the accurate grasp of channel contacts, making it difficult to achieve company goals. The operator wants to use the rapid modeling and application support capability of the machine learning platform to improve marketing efficiency and effectiveness.

Construction plan

AsiaInfo Technologies uses AISWare AI² to deploy a machine learning platform for the operator. It is a model training and application platform based on big data platforms and machine learning technologies. It supports the rapid implementation of data mining models through wizard-based modeling and intelligent assisted modeling. Since the platform was launched, the original manual offline modeling method has been changed to the rapid modeling based on machine learning platform, such as: 4G unified price package migration model, 4G off-network monthly churn early warning model, whole network user churn early warning model, and low saturation user identifying models, upgrade complaint risk early warning models, etc.

Application effect

The machine learning platform was successfully launched for the operator in November 2017 and has continued to evolve. The platform has completed the deployment of multiple models and achieved the expected goals.

  • High marketing conversion

    The accuracy of user identification in each model has been improved. For example, in the application of the 4G unified price package migration model, the marketing success rate has increased by 170% compared to that of selected users based on human experience.

  • Guaranteed revenue growth

    The revenue growth is guaranteed through targeted user marketing which could reduce user churn and upgrade user packages. For example, through user package upgrade marketing, the average user ARPU value has increased by 37%.

  • High modeling efficiency

    The modeling cycle using the machine learning platform has been shortened by about 67% compared with that using the traditional manual method.