Correlation n-ptychs of Multidimensional Datasets

Abstract

Correlation analysis studies the predictive potential stored in attribute pairs of multidimensional datasets, which is subsequently summarized through a correlation matrix or visualized using a correlation heatmap. Both of these approaches are suitable for examining pairs of attributes and the correlation values between them but are difficult to use when working with a larger number of attributes. Datasets containing several dozen or hundreds of attributes are, in the era of big data, gaining importance rapidly and are often the basis for building analytical models using machine or deep learning models. This paper focuses on the visual representation of parts of multidimensional datasets that carry a significant part of the prediction potential - correlation n-ptychs, while we consider the values $2 \le n \ge 6$. The proposed concept is subsequently presented in a case study conducted on a dataset focused on renewable energy and weather conditions.