70 On-Demand

Inspiration

Keep it simple.

What it does

Predicts the next three months of demand for the products based on historical data.

How we built it

• Scikit learn model trains on historical data and predicts next three months unit sales
• Flask server to serve results from model to UI
• UI in Vue JS for visualization results additional to the generated csv with predictions

Challenges we ran into

• Model training time
• Feature engineering wasn't working
• Linear regression model had a very low R2 score, so we opted for

Accomplishments that we're proud of

• 31.09 MSE
• 81.72% R2 Score on Holdout Set
• Successful predictions exported as CSV

What we learned

• Technologies used
• Feature Engineering

What's next for 70 On-Demand

• Create temperature and sunshine conditions to be able to predict demand at any point in the future
• Random forest is our baseline, so we expect neural network will outperform with the features used.

Built With

• flask
• scikit-learn
• vue