Gaussian Sampling Approach to deal with Imbalanced Telemetry Datasets in Industrial Applications

Practical implementation of data analytics in industrial environments has always been a problematic area because of data availability and quality. In this paper, a Gaussian sampling methodology is proposed to address the problem of imbalanced telemetry datasets that is one of the root causes that make modelling less reliable. By generating subsets that achieve homogeneous density distributions this problem is addressed. By comparing the impact of this method with the baseline case of random sampling, this paper aims to address this problem and propose a practical solution. A case study based on an industrial cooling device is used to assess and illustrate the proposed approach.