lab-machinelearning

Machine Learning experiments (DQR + predictions) on an example dataset.

Required environment

• Python v3.4+
• PIP

Install dependencies

pip install pandas numpy plotly sklearn scipy

How to use

This experiment learns from daily and hourly records of bike rentals over 2 years, and tries to predict the amount of bikes rented on a given day (or given hour), knowing the weather and other features.

1. Normalize input data

Our algorithms are fed a raw data extract. We need to do some normalization work before we can do any stats on it:

./1-normalize-data.py

2. Analize data quality

In order to perform a relevant prediction, we need to run a Data Quality Report (aka DQR):

./2-generate-dqr.py

3. Generate the prediction model

We can now generate the prediction model (from a % of the input data), and test prediction quality (on the remaining % of the input data):

./3-generate-model.py

Preview

Prediction quality report graph:

Copyrights

Refer to the README.txt file in the ./data directory for input data licensing notes.