Research And Application Of Multidimensional Data Anomaly Detection Method

Anomaly detection has been widely used in real life.However,there are still some problems in the current relatively common anomaly detection methods,which mainly focus on two aspects: one is the slow detection speed and low efficiency when dealing with massive or multi-dimensional data;Second,most detection only optimizes the description of normal samples,and does not optimize the description of abnormal samples,which may cause a large number of false positives or omissions.After analyzing some commonly used methods of anomaly detection,this paper uses the idea of fuzzy logic to perform fuzzy operation on the detection results of isolated forest algorithm,and proposes a method of anomaly detection for multi-dimensional data based on Fuzzy isolated forest,which solves the problem that different instances in multi-dimensional data have different degree of anomaly for different attributes,resulting in inaccurate detection.The main work is as follows:(1)The isolated forest algorithm builds trees by randomly selecting attributes to form the isolated forest.At last,it scores the traversal results of each sample and determines whether the sample is abnormal data by the value of the abnormal score.However,in practical application,the degree of anomaly of randomly selected attributes is different for each instance,and the detection inaccuracy will still exist if the anomaly score value of random attributes is generalized.According to the above problem is proposed in this paper an algorithm based on fuzzy isolated forest of multidimensional data anomaly detection method,starting from the multi-dimensional,considering each data and the nature of the subordinate relations between each dimension,using the membership functions of each dimension attribute of membership degree of monitoring results of judgment,and again by the membership degree of evaluation object and set relations of fuzzy matrix operations,a fuzzy evaluation results.Finally,a real data set of UCI dataset is used to compare the detection efficiency and accuracy of this method with other methods,which proves that the method proposed in this paper not only improves the detection rate of the algorithm,but also improves the comprehensive performance of the algorithm.(2)Fuzzy isolated forests algorithm is applied to the practical problems: according to the real campus identification data,analyzes the characteristics of college students' daily behavior,summarizes the influence the student's daily behavior produce abnormal six main factors,and finally designed a students' daily behavior anomaly detection model,through the experimental results and the experimental results of validation,proved in this paper,the proposed multidimensional data anomaly detection method based on fuzzy isolated forests effectiveness in practical application.The experimental results show that the multidimensional data anomaly detection method based on fuzzy isolated forest proposed in this paper has higher detection ability,can effectively analyze data,and has higher accuracy and computational efficiency.