Hot-dogs predictor

Inspiration

Use real-data to make predictions

What it does

It predict the quantity of hotdog sales per day in order to have a robust cook plan in order to avoid waste.

How we built it

Using the provided data, I sought to understand the relationship between features and how they influence the quantity needed to be forecasted. Before forecasting, I looked into these relationships and created some visualizations to analyze the data. For forecasting, I tried to apply some machine learning algorithms to find some patterns in the data in order to make predictions without being explicitly programmed to do so. I trained 3 CatBoostRegressor models, an algorithm which is known for dealing greatly with categorical data. CatBoost is a machine learning algorithm that uses gradient boosting on decision trees. Tried also other models: a simple Multi Layer Perceptron and a time series analysis method -> ARIMA.

Challenges we ran into

• feature selection, extraction
• Dealing with categorical variables
• finding an good forecasting model

Accomplishments that we're proud of

• build a model capable to forecast -learned a lot of new things

What we learned

• eda, feature extraction
• forecasting, timeseries

What's next for Hot-dogs predictor

• better data
• better algorithms

Built With

• matplotlib
• numpy
• pandas
• python
• sklearn
• tensorflow