Machine Learning Classification Techniques to Predict Directional Change of Energy Prices Using High Dimensionality Reduction

Abstract

In today's economy, energy is an essential commodity. Every individual and business use energy for their needs. Electricity is the most common form in which energy is consumed. Thus, accurately predicting electricity prices could aid businesses in planning their finances and logistics and have a better long-term vision of their company. In this paper, the next day directional change of the electricity prices of the German and Austrian areas of the European Energy Exchange (EEX) wholesale market is predicted based on several parameters, including the daily Phelix index, the volume of trade, coal prices, Title Transfer Facility (TTF), wind power production, and many others. High Dimensionality Reduction techniques (Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA)) are used in conjunction with Machine Learning (ML) classification algorithms; Support Vector Machines (SVM), and Artificial Neural Networks (ANN), in particular. The software employed for this research was Python 3, used on Google Collaboratory. The maximum forecast accuracy achieved by our model was 75.00%.