Study Of Abnormal Call Behavior Decection In Big Data Environment

With the continuous development of the communications industry,the majority of users began to gradually enjoy a variety of communications services.However,more and more malignant abnormal calls are followed.A large number of groups or individuals harass the target population by means of telecommunications fraud,marketing or personal attacks,so that ordinary users can not communicate normally,resulting in poor user experience.This behavior seriously affects users' normal use of network resources and deteriorate the operator's brand image.Therefore,the research on the recognition of abnormal call behavior has very important practical significance.Because the daily call record data of the telecommunication company is above ten million,the recognition of abnormal behaviors requires the analysis of the huge amount of call data.However,it is very difficult to store and process huge amounts of data.Only by designing reasonable distributed system and using data mining technology can we analyze the business data of this order of magnitude and realize the detection of abnormal behaviors.This article first describes the background of abnormal call detection,research status quo,and describes the main work of the paper.Then learning the relevant technology,learning more about data mining algorithms,and focusing on data processing and the classification of random forest algorithms related content.In this paper,the random forest algorithm is applied to the field of abnormal call detection.Aiming at the unbalanced data of abnormal call and normal call in the phone bill,a KSR solution is proposed.That is to say,the K-Means clustering algorithm is used to down-sample the majority samples and the SMOTE algorithm is used to sample a few samples,which can make the data set more balanced.Then,the random forest classifier is used to train the samples,and finally the model is verified and analyzed.Experimental results show that the classifier trained by this scheme has a very good effect on the indicators such as the false positive rate and the success rate.At the same time,in order to store and deal with huge amounts of bill data,we designed an abnormal call behavior analysis system.The Elasticsearch cluster is used to make full use of the power of the cluster for high-speed computing and storage.The article designs data acquisition,data preprocessing,data storage and analysis modules.And the article mainly expounds the data preprocessing module,introduces its main flow and proposes a hierarchical file processing method to reduce the memory footprint.Through testing the system,the results show that the designed abnormal call detection system has a good effect.