spark-project

This project involves:

• Collecting raw data from numerious devices installed in several buildings of University of Waterloo.
• Crunching the motion sensor raw data in timeseries format categorized with respect to devices.
• Find the most similar devices through motion sensor data using Cosine Similarity.
• Predicting occupancies within the facilities accross different prediction horizons using specific machine learning models.

The following figure shows the prediction results of an algorithm called persistence which was used in the project.