Brainstorming on Dataset Reduction from an Heuristic Bioinspired Green Computing Approach

Artificial intelligence has become essential in our daily lives, and over time we have become more dependent on it. As this tool becomes more precise, energy consumption increases, and this has become a concern for sustainability. Green learning is proposed as a solution to address these concerns. In this work, we propose to perform PV power generation prediction by applying training data size reduction using a machine learning approach performed with evolutionary techniques. In this way, the processing time of the classifier is accelerated without reaching a significant loss on the effectiveness of the classifier. The data reduction performed achieves a reduction of 17.13%, and the effectiveness is reduced by 1.47%. With respect to the initial problem, reducing the training time reduces the carbon footprint produced by energy consumption.