A Method For Finding And Eliminating Outliers In Data Mining Classification Tasks

Data mining is technique used to extract interesting data patterns from datasets. A vital piece of data mining is outlier detection, that detects objects exhibiting behaviour different from the rest. One common issue in data mining field is identifying outliers from a set of patterns. An outlier can significantly affect the result. Thus, in order to avoid skewing, outliers must be eliminated or resolved prior to the analysis. Data mining seeks to uncover patterns within data that deviate from expected behaviour in order to detect outliers. In this paper, we have suggested a system that uses the widely-used statistical technique known as IQR to find out outliers and use the Winsorizing method to handle them. Under this architecture, the pre-processed dataset is used to create RBFN that is prepared by teaching a learning-based optimisation model. As a result of this study, our proposed framework may serve as a useful substitute tool for data mining classification tasks where outliers preprocessing is required.