Improvement Research Based On KNN Algorithm And Its Application In Data Classification

The nearest neighbor K(KNN,K-Nearest Neighbor)classification algorithm is one of the simplest methods in the data mining classification technology.Because of its simplicity,it has been widely used in many fields.However,when the sample size is large and the attribute attributes are large,the efficiency of KNN algorithm classification will be greatly reduced.In this paper,we propose an improved KNN algorithm and compare it with the traditional KNN algorithm.This algorithm does not directly predict the value of the response variable,but reduces the range of the maximum likelihood of the response variable,and then interpolates to give the output.In the preprocessing step,we divide the data into layers,and run the search response with the most possible partition.It uses a single parameter K,which is the same as the traditional KNN algorithm,and surpasses the conventional technical methods in various datasets currently shown by experiments.This paper proposes a novel and efficient and has a stray resistance KNN regression algorithm based on clustering,CLUEKR algorithm firstly find the query point,rather than direct search recent data in the whole data set,and then find which cluster.This algorithm firstly in the preprocessing steps to hierarchical clustering of data,then execute the starting from the root node of the hierarchy of recursive search,search in the hierarchy of the current node,select a cluster between child nodes,and then used a recursive search.Finally,the k nearest neighbor of the query point in the cluster is found and the weighted average of its response variable is returned.This paper also proposes that the CLUEKR algorithm can be used to modify the classification task.In addition,this paper puts forward a weighted k-nearest neighbor algorithm based on class,according to the example,the classification of the neighbor domain of the regular k-nearest neighbor classifier is given,and the weight is assigned to each class.The algorithm considers the class distribution near the query instance and ensures that the assigned weight does not adversely affect the outliers.A thorough experimental study of the methods proposed in several real world datasets has proved that the method is better than the most advanced one.Finally,this paper combines the weighted k-nearest neighbor algorithm of class and CLUEKR algorithm,and proposes an efficient and accurate KNN classifier considering the data nature.