Proposal of data preparation model for Big Data analytics in painting process

Abstract

This paper deals with a painting process in a car company, focused on selection, obtaining and preparation of the data required for data mining analysis. The painting process is a very complex one with various parts, producing huge volumes of data, which are not only heterogenous, but oftentimes not at all time-synchronised and/or missing a common identificator. Thence, the data acquisition part is crucial in the analysis process, and is required to be well-thoughtout and documented. The first part of the paper introduces applications of big data analysis methods in general and with focus on industry, followed by a description of the selected process in the real company and identification of the goal of desired analysis. The second part focuses on acquisition of the required data, therefore proposes the various data sources within the company, the selection process, obtaining of the data (samples or, preferably, in real time), and integration of the obtained. The third part proposes a model for pre-processing and transformation process, with closer look upon the problems and issues specific for such heterogenous data volume. The fourth, final part summarises the results of data preparation and drafts the further analysis of the process.