iKhushPatel/Predicting-Bike-Sharing-Patterns

Machine Learning | Predicting Bike Sharing Patterns

Predicting Bike-Sharing Patterns

The first Udacity Project from the Deep Learning Nanodegree. Using only Numpy to implement forward and back propagation to predict Bike-Sharing Patterns.

Setup:

1. Create a conda environment: Make sure you have installed: numpy matplotlib pandas jupyter notebook

conda create --name deep-learning python=3
conda install numpy matplotlib pandas jupyter notebook

2. Activate your environment:

• Windows:

activate deep-learning

• Mac/Linux:

source activate deep-learning

3. Run the notebook server in the root directory of the project:

jupyter notebook

• The server is located in: localhost:8888

4. Open the Jupyter notebook called: Your_first_neural_network.ipynb

• The data comes from: UCI Machine Learning Database

Results to expect:

Training and validation loss:

Predictions after training:

Conclusions

The predictions are acceptable but in the December the model overestimes bike ridership because it hasn't had sufficient holiday season training examples.