Model Training and Evaluation

This repository contains the code for training and evaluating a machine learning model to predict financial outcomes based on various features, including text data. More processing code will be uploaded in due course.

Table of Contents

• Requirements
• Usage
• Code Structure
• Arguments
• Citation

Requirements

To install the required libraries, run:

pip install -r requirements.txt

Usage

To run the script, use:

python main.py <<insert flags>>

For example:

python main.py  --no_text False --no_text_features 'all' --only_text False

Code Structure

The main script is structured as follows:

• Imports: Necessary libraries and modules.
• Argument Parsing: Parsing command-line arguments.
• Helper Functions: Additional functions for text processing and data formatting.
• ModelTrainer Class: The core class for training, evaluating, and saving the model.

Arguments

The script accepts the following command-line arguments:

• --no_text: Boolean, default is False. Indicates if no text features should be included.
• --no_text_features: String, default is 'all'. Specifies the type of features to exclude when --no_text is True.
• --only_text: Boolean, default is False. Indicates if only text features should be included.

Citation

If you use the code in this repository, please cite the following work:

@misc{drinkall2024traditionalmethodsoutperformgenerative, title={Traditional Methods Outperform Generative LLMs at Forecasting Credit Ratings}, author={Felix Drinkall and Janet B. Pierrehumbert and Stefan Zohren}, year={2024}, eprint={2407.17624}, archivePrefix={arXiv}, primaryClass={q-fin.RM}, url={https://arxiv.org/abs/2407.17624}, }