This project is based on this repo and this associated blog post published in MLearning.ai which goes into more detail on the methodology presented. The main idea for this project is to download a lot of data that we will process and train a model on, which we will then later use to predict future recessions. For that we will use 3 different methods, logistic regression, random forest and XGBoost.
This project is split up into two parts. The data acquasition and the model creating and prediction part.
- Full Potential - this notebook downloads, preprocesses and saves all of the used data in a local csv file which will be read by the other part of the project. In here we download a lot of data from FRED and nasdaqdatalink. To further understand why we chose the follwing indicators for our project, you should take a look at Daniels blog post.
- Model Tuning - this notebook is where the magic happens and our models are trained and executed. We will also take a closer look at the preformance of the models and compare them which each other. At the end we also try to predict in a monthly basis instead of fiscal quarter basis to try to get a prediction as close as possible to the current date.
conda create -n aoec -f environment.yml
conda activate aoec