I’m excited to share a demo app I’ve made that lets you train and evaluate a L2 regression (Ridge) model for predicting the interest rates of loans based on various factors. It lets you see how well a model performs (actual vs predicted plot), as well as compare the importance of these features.
In this app, we query data from a Snowflake Data Warehouse. If you are not familiar with Snowflake, you can check this out, or watch this intro video. In the bottom left, you can see the SQL query that is automatically generated from the controls; this query is sent to the Snowflake server, and a subset of the 1.5M loan records is returned as a dataframe. All this can be done in a few lines in Python, thanks to the official Snowflake Python APIs, as well as SQLalchemy/pandas.
One really nice thing about Snowflake is that we don’t have to worry about configuring access to databases and computation resources, so we can solely focus on composing the right queries, building the app and modeling the data.
You can find the source code in our Github repository, which contains all the instructions for running your app locally.