Research On The Recommendation Of Mobile Phone Tariff Package Based On Idata And (?)str(?)m Algorithm

With communications well-developed,tariff packages selection has become today's topic of concern for everyone.Choosing a suitable tariff package has already become an important guarantee for saving personal communication costs.However,most consumers passively accept the package recommended to them by operator and do not realize that it is not suitable for them.This can easily lead to a waste or a lack of package.For users who want to change their package,there is a lack of tools and methods to help them choose a suited package based on personal historical consumption data.Therefore,this article combines package recommended method and iData technology to study iData mobile phone tariff packages recommendation,which has a high value of practical application.Everyone's communication data is unique and individual,which is so-called"personal" data,iData,,"my data",or "little data".The package recommended method studied in this paper is mainly based on individual data to analyze the individualpackage usage and recommend packages to consumers.In view of the research status of communications industry and application requirement in recent years,mobile phone tariff package recommendation algorithm based on iData and Astrom algorithm is given The algorithm includes two part:A bill factor predicting model based on improved Astrom forecasting algorithm and a mobile phone package recommendation algorithm based on bill factor prediction Prediction models,which are based on traditional ARIMA model,are established espectively for bill factors such as local call duration,long-distance call duration(7:00-23:00),long-distance call duration(23:00-7:00),roaming call duration,and data traffic.By adding error correction factors to the Astrom prediction algorithm,an improved Astrom prediction algorithm is given to achieve multi-step forecasting of bill factor.It can reduce the prediction error of the traditional ARIMA model to a certain extent and improve the accuracy of multi-step prediction Taking the predicted value of each factor as the input condition,two mobile phone package recommendation algorithms based on billing factors are proposed to recommend the appropriate package for users.One of them is the fixed package recommendation algorithm for fixed packages given by the operator,and the other is the optional package preferences algorithm for designing a reasonable combination package with custom value of each bill factor.Experiments are conducted to verify the bill factor prediction model and mobile phone package recommendation algorithm based on the actual consumption data of a mobile communication business office in Dalian Experimental results show that the prediction error of the bill factor value is within an acceptable range,and each user can get a suitable package according to the prediction result.In addition,in this paper a mobile phone tariff package recommendation system are designed and most function modules and implemented.Through this system consumers can intuitively search their own consumption data and suited packages which can help them save communication expenses,and communitction perators can also recommend suited packages to consumers so as to reduce the consumers' offline rate.