Mcgill_fiam

Instructions to run the code

This repository contains a series of scripts for data preprocessing, feature engineering, and feature importance analysis.

Getting Started

Prerequisites

• Python 3.7 or higher
• pip (Python package installer)

Installation

1. Install the required packages:

   pip install -r requirements.txt
   

2. Create a new directory raw_data and add the following files to it:
   • factor_char_list.csv
   • hackathon_sample_v2.csv
   • mkt_ind.csv

Running the Analysis

Follow these steps to run the complete analysis:

1. Data Preprocessing:

   cd 01-Data_Preprocessing
   python preprocessing_code.py
   cd ..
   

2. Feature Engineering:

   cd 02-Feature_Engineering
   python feature_engineering_code.py
   cd ..
   

3. Feature Importance:

   cd 03-Feature_Importance
   python feature_importance_code.py
   python feature_selection_code.py
   cd ..
   

4. Causal Discovery:

   cd 0X-Causal_discovery   
   python discovery.py
   cd ..
   

5. Predictor:

   cd 04-Predictor
   python train_AlphaSignals.py
   cd ..
   

6. Backtesting

   cd 06-Backtesting
   python backtest_parallel.py
   cd ..
   

7. Chain of Thought Zero Shot Features Download the dataset from here and put it in the datasets directory. Login into Hugging Face and download the model here.

   cd 0L-CoTZeroShotFeatures
   python create_dataset.py
   python llama-3.2-3B-Instruct-Inference-READABILITY-SCORE.py
   python llama-3.2-3B-Instruct-Inference-RISK_FACTORS.py
   python llama-3.2-3B-Instruct-Inference-SENTIMENT-SCORES.py
   cd ..
   

Output

After running all the scripts, you'll find the following output:

• objects/FULL_stacked_data.pkl
• objects/causal_dataset.pkl
• objects/WEIGHT_SAMPLING.pkl
• objects/mkt_ind.csv
• objects/X_DATASET.pkl
• objects/Y_DATASET.pkl
• objects/predictions_0.csv to predictions_13.csv
• objects/predictions.csv -- Final predictions on the test dataset
• objects/prices.pkl -- Dataframe of the prices of the assets
• objects/signals.pkl -- Timeseries of signals from each stock
• objects/market_caps.pkl -- Dataframe of market caps of the assets
• objects/stock_exret.pkl -- Excess returns of the each stock

The feature importance scores from MDI and MDA are:

Additional Information

For more details on each step of the process, please refer to the comments and docstrings within each script.