Optimizing Correlation Measure Based Exploratory Analysis

Abstract

Exploratory data analysis refers to the existing data to explore under the assumption of less as far as possible, through drawing, tabulation, calculation methods to explore characteristics of data structure and regularity of a kind of analysis method. However, exploratory data by calculation method is a very general method to find the key of data. In this paper, we introduce a correlation measure for exploratory analysis based on maximal information coefficient. First, we briefly introduce the traditional data analysis methods and features, expound the necessity of exploring analysis and content. Then, correlation measurement which used commonly are expounded, summarized their characteristics and models. Therefore, we propose a weighted measure based on Maximal Information Coefficient to improve effectiveness of exploratory analysis. Then we get eigenvalues of the Maximal Information Coefficient and Pearson correlation coefficient in linear and nonlinear function of plus noise. Finally, explore analysis display visualization of the results by test dataset, emphasis direction of further research.