Stock-Treadliner

Founders: Jasper Liu, Ngoc Vo, Ngoc Tu, Thuy Linh Pham, Nathanael Garcia, Nicholas Magtangob

Description: Given the stock dataset with daily OHLCV (Open, High, Low, Close, and Volume) values, we will use a machine learning pipeline to predict multiple stocks’ next-day logarithmic returns, defined as $y_{t+1} = \dfrac{log(close_{t+1})} {close_{t}}$, for the upcoming year.

Get Started

Read this final report document to get a better understanding of the project

You can run the example notebooks directly from the /experiment directory to test the project workflows. However, for the scripts inside /backend, make sure to run them from the project root, e.g.:

python ./backend/evaluation.py

Acknowledgement

This project uses:

1. yfinance Stock Market Dataset (© Ran Aroussi, licensed under the Apache License 2.0)

Project Structure

1. /experiment: where the jupyter notebook and testing files live
2. /backend: where the machine learning pipeline and backend server live
3. /frontend: where the user interface lives
4. /data: where the processed and feature datasets live
5. /model: where meta-information about fine-tuned and ensembled models lives
6. /graph: where evaluation plots for each model live