Research And Implementation Of Database Abnormal Point Mining Method Based On Neighbor Relationship

With the rapid development of database storage,computer science and data acquisition technology,large amounts of raw data have accumulated in databases in recent years,Therefore,How to extract valuable information from massive data quickly and accurately is a hot topic in the field of computer science.Data mining is a technology for extracting valuable knowledge from large amounts of data.Hence,database outlier mining refers to the process of discovering knowledge or information with high value from the database contains mass data through certain methods.This paper focuses on the neighbor relationship between data points,and proposes a database outlier detection named NLOF(Neighbor-Local Outlier Factor)based on neighbor relationship.Compared with the traditional outlier detection algorithm based on density,this algorithm has achieved visible improvement in precision,cost of time,etc.The major improvements of the algorithm are be listed as below:1.Putting forward a filter method based on neighbor relationship to preprocess the data and grid ridge point.2.Using information entropy to reduce the data dimension and weight the properties.3.Optimizing a neighbor query method based on grids neighbor relationship,and reducing the computation complexity of finding the data neighborhood.This paper verifies the algorithm through experiments,using synthetic data,data set of wine quality and UCI(University of California,Irvine)adult data.Experiment results show that:this method has achieved improvement in the aspects of checking efficiency and precision compared to LOF(Local Outlier Factor).