deep-time-series-forecasting-models-for-stock-price-prediction

Intro

A collection of prominent and SOTA time-series forecasting models, including deep, uni-variate, multi-variate and ensemble models.

Installation

To get started, you'll need Python and pip installed.

1. Clone the Git repository

git clone https://github.com/anaeim/deep-stock-price-prediction.git

2. Navigate to the project directory

cd deep-stock-price-prediction

3. Create directories for the stock data and the trained models
   You can download Apple (AAPL) Historical Stock Data from this Kaggle web page and Tesla (TSLA) stock data from this Kaggle web page.

mkdir data
mkdir model_keeper

4. Install the requirements

pip install -r requirements.txt

Models

• uni-variate LSTM
• Prophet
• NeuralProphet
• multi-variate LSTM
• XGBoost
• LightGBM
• ensemble models:
  • XGBoost + multi-variate LSTM
  • XGBoost + LightGBM
  • LightGBM + multi-variate LSTM
  • XGBoost + LightGBM + multi-variate LSTM

Training

python train.py --dataset AAPL \
    --ml-model ensemble_XGBoost_lstm_multivariate \
    --test-size 0.35 \
    --time-stamp 100 \
    --epoch 1000 \
    --batch-size 84 \
    --enable-save-model

The meaning of the flags:

• --dataset-path: the directory that contains the dataset
• --ml-model: the Machine Learning (ML) model that we use for time-series forecasting
• --time-stamp: the time stamp to create windowed datasets
• --test-size, --epochs, and --batch-size are the proportion of the dataset to include in the test split, number of epochs and number of training examples in each iteration of the model, respectively.
• enable-save-model: to save the trained model into the model_keeper directory

Prediction

python predict.py --dataset AAPL \
    --ml-model ensemble_XGBoost_lstm_multivariate