Wine Data App Part 2

Data Analysis

We decided to do the data preprocessing and selection using Jupyter Notebook. Similar to an R workspace, Jupyter Notebooks have the ability to easily integrate Python packages for data analysis. Python code can be integrated into a markdown file and run directly in the notebook to produce results instantly within the same markdown file to be exported if desired. In our case, we finally decided on using a random forest model to fit our data. The model was exported to a binary file using the Pickle package to be used in our back end API.

App

We used the Python API framework Flask to create our app. Our pickle model was loaded into the app and we rendered an HTML page with simple Bootstrap css styling. The page consists of input boxes which allow a user to input custom values for the chemical properties of a wine. Submitting the form will then give the user a quality score predicted using our random forest model.