This repository was created as part of a course capstone project for GalvanizeU M.S. in Data Science for 6007- Data Enginering in Spring 2016. It will be periodically extended to experiment with machine learning on sensor data.
It contains parts of an experimental smart meter disaggregation + energy savings recommender application with the goal to personally learn key data processing. The application uses Postgres and Redshift for storage, Kinesis for handling streaming data, Spark for data processing, and Graphlab for an online Bayesian checkpoint model for anomaly detection.
Update (2016-05-11): Adding in experiments with demand forecasting using Recurrent Neural Networks via Keras.