Synthetic Ground Truth Generation of an Electricity Consumption Dataset

The training of supervised Machine Learning (ML) and Artificial Intelligence (AI) algorithms is strongly affected by the goodness of the input data. To this end, this paper proposes an innovative synthetic ground truth generation algorithm. The methodology is based on applying a data reduction with Symbolic Aggregate Approximation (SAX). In addition, a Classification And Regression Tree (CART) is employed to identify the best granularity of the data reduction. The proposed algorithm has been applied to telecommunication (TLC) sites dataset by analyzing their electricity consumption patterns. The presented approach substantially reduced the dispersion of the dataset compared to the raw dataset, thus reducing the effort required to train the supervised algorithms.