USIO Energy Data Science Exercise

You will use historic energy usage data from a number of households to forecast their future usage.

Data

usage_train.csv contains historic data in a single file. The columns are:

• id: a unique household ID
• datetime: datetime of the half-hour interval for which usage is recorded
• usage: usage for the given household and half-hour (in kWh)

usage_test.csv is similar to usage_train.csv, but is missing the usage column. We would like you to predict usage for each row (household/half-hour) in usage_test.csv.

Submission

We are interested in how you approach an open-ended machine learning problem. We encourage plotting and data exploration, as well as readable and re-usable code.

• use python 3
• fork this repository
• commit all your code (including Jupyter Notebooks, exploratory work etc.)
• submit a pull request when you are finished
• please make sure the code includes two files:
  • train.py should read the training data, perform any steps needed, and save a single file containing the trained forecasting model (e.g. model.ext)
  • predict.py should read the test data (usage_test.csv) and the model file (model.ext), and save predictions in usage_test_predictions.csv.
• feel free to organise additional files as you see fit
• please update QUESTIONS.md and try to answer questions therein.
• please include a description of external libraries needed to run your code

Please try to not spend more than half a day on this exercise! Have fun!