Week 11: Working on the data

For this week, the team worked on multiple areas. Articles about consumption optimization were briefly examined since optimization algorithms will be of use in the future; thus, an investigation was needed.

Then the data about the electrical consumption required an expansion so that we would have more information regarding households that could be used as features. With the data expansion, we aimed to obtain information about households' locations, the dates of holidays and weekends, temperature, humidity, and other characteristics such as residential electricity consumption.

It was also needed a way of grouping the data to be better studied. This was achieved using agglutination in one-hour periods was adopted. This periodicity sufficiently represents the data and allows for less processing for the machine learning and planning algorithms. In addition to data treatments, some algorithms that forecast household consumption behavior for future moments were tested, such as the Support Vector Regression (SVR). Although it requires adjustments in the models and configurations to improve their predictions.