Rearch On Traffic Prediction Based On Frequent Pattern Mining Algorithm

Prediction has been one of the major purposes in big data analytics,and has shown great impacts in many fields.In the implementation of resource scheduling of the base station,it is not always necessary to predict the actual traffic accurately,but the traffic status.This is very important for the base station's sleep policy,bandwidth,power and resource block control.In terms of domestic and foreign research on the traffic forecast based on wireless cellular network,yet there is no relevant algorithm can solve the problem of traffic status forecast reasonably.Because in a continuous short period of time,the transition of traffic state is very small,so in most cases there is little difference between the next moment's traffic and the current moment.This means that when quantify traffic to a limited number of states,the adjacent moments will be in the same state.And then if we use the principle of frequent pattern mining algorithm of maximum support to predict,it will lead to the future state of the forecast are the same with the current state,showing the state of continuous coverage of the phenomenon which can not achieve accurate prediction.Therefore,when applying frequent pattern mining to prediction,using the principle of maximum support to find the matching pattern is not suitable for time series prediction.In order to solve these problems,a frequent pattern mining method is proposed in this paper,called Frequent Pattern Mining-Matching(FPM-Matching),for traffic prediction with spatiotemporal model.The main contents in this dissertation are summarized as follows:1.Frequent Patterns Mining for Wireless Cellular DataIn order to explore the association rules of wireless cellular network data and mine the frequent patterns of cellular network data,different types of traffic data are analyzed.After analyzing the cross-correlation and auto-correlation of base's traffic,an efficient FPM-Matching algorithm is proposed,which improves the prediction accuracy and efficiency compared with other algorithms.The results presented in this paper are based on analysis of measurement data collected at two different locations in Nanjing and Hong Kong.Association rules are established based on the correlation of the data.The sequence satisfying the threshold condition,which is screened by the principle of support and confidence,is called frequent sequence,and the frequent sequences are stored by using the tree structure.So that the frequent sequences build up the tree structure according to the hierarchical and node progressive way.In this case,each frequent sequence will have a position coordinate in the structure.2.Research on traffic prediction based on frequent pattern miningThe 15-day base station traffic data is organized into a two-dimensional matrix.The number of rows in the matrix represents the total number of base stations and the number of columns in the matrix represents the total number of moments.In the process of forecasting,a method of maximizing the matrix to achieve frequent pattern matching is proposed,in which maximizing the matrix means using the current moment as the last column of the matching matrix,expanding the matrix continuously to the left until the matched matrix meet the following two conditions:First,the matrix belongs to the frequent pattern tree.Second,all the extended matrixes of the matrix do not belong to the frequent pattern tree.We mine and construct frequent pattern tree from wireless cellular network data using FPM-Matching algorithm.Through the frequent pattern sets of tree structure,it can determine the position of prediction mode quickly,obtaining the candidate sets.The evaluation results show that the FPM-Matching algorithm has the advantages of accuracy and effectiveness in prediction based on spatiotemporal modeling.The optimized matching pattern method improves the accuracy of the prediction results significantly.In terms of time complexity and accuracy,the FPM-Matching algorithm has obvious advantages,which can reduce the algorithm running time and the input of the algorithm reasonably.3.Research on 3D frequent pattern mining and traffic forecasting based on spatio-temporal modelIn the implementation of resource scheduling,some areas need to analyze and predict traffic data.According to the aggregation processing analysis of the grid data of the base station,the cross-correlation of aggregated traffic data is obviously higher than that of the non-aggregated data.To explore the effectiveness of FPM-Matching based spatio-temporal traffic modeling in wireless cellular networks,in this paper,traffic data of base station with latitude and longitude information is meshed from three aspects of traffic,geo-spatial location and timing,and the traffic volume of base station is aggregated into grid traffic.The XY-axis corresponds to the spatial location of the grid and the Z-axis corresponds at the moment,the time series of base station traffic is organized into a three-dimensional matrix.The FPM-Matching algorithm is used to construct frequent patterns in 3D matrices and the accurate prediction of space-time traffic data is realized by using the constructed 3D frequent patterns.