Overview

This repository contains the refined implementation of my thesis which investigates the performance of machine learning models in forecasting key macroeconomic indicators.

Setup

This project uses uv to manage the virtual environment and dependencies. Follow the offical guide to install uv on your device. Once installed, you can initialize this project with the following steps.

1. Clone the repo:

   git clone https://github.com/AndyYTHsiao/ml-macro-forecasting.git
   cd ml-macro-forecasting

2. Sync the environments:

   uv sync --dev

   This command automatically creates a virtual environment and installs all dependencies specified in pyproject.toml.

Quick start

A sample implementation is available at run_experiments.py.

# With uv
uv run python -m src.run_experiments

# Without uv
python -m src.run_experiments

If you want to tune and train the model on your own, you can follow the following steps.

from pathlib import Path
from src.core.tuning import Tuner
from src.core.training import Trainer

# Assume you have prepared the training and test datasets

# Basic setting
model_type = "svm"
framework = "sklearn"
task = "regression"
n_trials = 100
cv_splits = 3
output_dir = Path("./outputs")

# Model tuning
tuner = Tuner(
    model_type=model_type,
    framework=framework,
    task=task,
)
best_params = tuner.tune(
    X_train=X_train,
    y_train=y_train,
    n_trials=n_trials,
    cv_splits=cv_splits,
)

# Model training
trainer = Trainer(
    model_type=model_type,
    params=best_params,
    framework=framework,
    task=task,
)
trainer.train(
    (X_train, y_train),
    (X_test, y_test),
)

# Model evaluation
trainer.evaluate(
    X_train=X_train,
    y_train=y_train,
    X_test=X_test,
    y_test=y_test,
    rounding=4,
)

# Save results
trainer.save_results(output_dir / framework / model_type)