This project attempts to create a Dendritic LSTM model for stock prediction. The download.py script dynamically creates the dataset used for this project. The download script attempts to get as much data as possible from companies traded on the NASDAQ (The script consistently gets ~3k companies in around ~30s). Unfortunately, due to starting this project on Saturday (1/17/26), no hyperparameter sweeps were performed.
Team:
- Vishwesswaran Gopal - First Year Engineering Student @ Purdue - Linkedin - gopal21@purdue.edu
LSTM models are commonly used for natural language tasks, such as machine translation and sentiment analysis; speech and audio processing, like speech and command recognition; time series forecasting, such as financial and weather forecasting; anomaly detection, i.e., fraud detection, and many more use cases not listed here.
Due to the importance and abundance of LSTM models, assessing whether LSTM models perform better under dendritic optimization remains crucial as dendritic optimization may allow these models to become cheaper and more efficient, thereby saving money, resources, and time.
Thus, this projects attempts to apply dendritic optimization to to a common use case of LSTM models, stock prediction, to determine the viability of dendritic LSTM models.
Installation:
git clone https://github.com/VG-Fish/Dendritic-Stock-Prediction-Model.git
cd "Dendritic-Stock-Prediction-Model"
pip install uv
uv sync --locked
Run:
# For dendritic LSTM model
uv run dendritic_main.py
# For normal LSTM model
uv run main.py
# To get more info about the parameters you can pass into the main scripts, run:
uv run <CHOOSE_MAIN>.py --helpWill be shared soon.