The main objective of the project is the creation of a machine learning model able to calculate accurate predictions of future breakdowns for both public luminaires (from now on lights) and electric panels (eboxes). The repo contains the whole pipeline, from raw data preprocessing to training the models and making predictions. The code has excellent comments at each step, ensuring a clear understanding of each module.

The model will take as input information about the breakdown history from the last 5 weeks of a light or an ebox together with context information such as electric readings data and weather conditions. The output is the probability of breakdown in the next 4 weeks.

The code has four main functionalities that can be combined depending on the necessities of the user:

• Meteo web scraping: Gather the meteorological information from a certain municipality over a specified time period using web scraping.
• Data preprocessing: Pipeline to do the necessary data preprocessing of the raw breakdown data, the raw readings data and the scrapped meteorological data.
• Model training: Training of the machine learning models using the prepocessed data.
• Doing predictions: Gather data from the last 5 weeks, preprocess it and make predictions using the trained models.