FinCLASS

Code release for the paper titled Modeling Financial Uncertainty with Multivariate Temporal Entropy-based Curriculums (link), accepted at the UAI 2021 conference as a full paper.

Environment & Installation Steps

Create an environment having Python 3.6 and install the following dependencies

pip3 install -r requirements.txt

Contents

• data_preprocessing_code - Contains all the scripts for preprocessing the data, and steps to run, information about preprocessing, etc.
• difficulty_score_code - Contains the code and the steps to run for computing the difficulty score defined in FinCLASS framework.
• model_training - Contains the code for training THA-Net both with and without using the curriculum generated using FinCLASS.

Data

Find the US S&P 500 data here, and the China & Hong Kong data here.

Steps

1. Preprocess the data using the scripts in data_preprocessing_code, following the instructions mentioned there
2. Calculate the different complexities using the scripts in difficulty_score_code
3. Train the model using the scripts in model_training, following the instructions there