Analysis And Research On The Characteristics Of Mobile Internet User Traffic Services

Mobile Internet is the main implementation of next generation communication technology.With the rapid development of China's communication technology,the scale of users has expanded and business traffic data has increased dramatically,with the consequent complexity of demand and the urgent need to improve user experience.Achieving an accurate portrayal of traffic business behaviour for different user groups can effectively improve the efficiency of operators for customised services.As there are large individual differences in user data and strong changes over time,this paper conducts user analysis and prediction research on the basis of this problem for traffic data,and the specific research work is as follows.(1)Research on the distribution pattern of user traffic usage time.A Hadoop-based big data platform was built,and Spark was applied to pull the source data stored in Kafka and the parameters in Redis to achieve data association,which was finally stored in Hive to achieve distributed storage and data pre-processing of a large amount of user data.The statistical analysis of traffic usage behaviour and user usage characteristics of the application from the user's perspective.The analysis shows that there is a clear long-tail effect in terms of user traffic usage,which plays a key role in the subsequent analysis,and in combination with the analysis of user access to application software,the behavioural preferences of users in terms of software categories are found.(2)Classification study of user types.Analyzed from both the time-domain characteristics of traffic usage and the characteristics of application software usage,firstly,factor analysis was applied to extract features from user traffic data,and the four public factors obtained corresponded to different time periods in a week,and five different user types were obtained through K-means clustering.The traffic usage behaviour of each type of user in different time periods differed significantly,and the correlation between the common factors was stronger and more interpretable than that of the principal component analysis method.The data was then downscaled using latent semantic method to obtain six types of users after clustering.Combined with the application software data to analyse their usage traffic behaviour,it was found that the users' software usage behaviour has a great influence on their traffic usage.(3)User usage traffic prediction studies.User traffic data can obviously be used as a kind of time series data and be predicted.An ARMA traffic prediction algorithm based on the daul tree complex wavelet transform is proposed.Since user traffic data has the characteristics of being unstable and strongly changing,the time series is pre-processed using the daul tree complex wavelet transform,and the decomposed series is selected for prediction by applying the ARMA model,and the prediction results of user traffic data are obtained after the daul tree complex wavelet inverse transform.Finally,it is compared with the traditional time series algorithm for the same traffic time series prediction and the error is calculated as an evaluation criterion,and the results show that the proposed algorithm has better prediction performance.The traffic usage behaviour classification and user traffic prediction under a large number of users is achieved,which can provide operators with traffic packages that better meet user needs for business formulation.