Gaussian Process Component Mining with the Apriori Algorithm

Abstract

Gaussian process models are a commonly used tool for model-based analysis of time series data. With growing database size, the difficulty to identify the most interesting insights in order to gain a deeper understanding of the data’s underlying behavior increases. To address this issue, we propose a novel approach for finding frequent kernel components efficiently. In this way, data scientists are empowered to focus their investigations on the most common parts hidden in a set of Gaussian process models. We show how to solve this task by means of frequent item set mining methods, which are capable of analyzing large databases efficiently. We provide evidence of our proposal with a first series of experiments, indicating that our method is capable of detecting frequent kernel components from Gaussian process models. Though this short paper can be thought of as a first preliminary approach towards analyzing Gaussian processes with conventional data mining methods, it simultaneously opens a novel research direction of Gaussian process mining at the intersection between machine learning and database research.

This research was supported by the research training group “Dataninja” (Trustworthy AI for Seamless Problem Solving: Next Generation Intelligence Joins Robust Data Analysis) funded by the German federal state of North Rhine-Westphalia.