Birdirectional-LSTM Based Stock Prediction

Introduction

This research was conducted for the purpose of analyzing and predicting stock price data based on Birdirectional-LSTM. About the top 10 KOSPI stocks for research We used crawling technology to collect stock information and company-related news information. It also consists of pattern capture technologies such as modified closing price, rolling and lagging, technical analysis, sentiment analysis, and Stacked Variational AutoEncoder.

Paper

Please refer to the paper (Korean) Drive Pages

Requirements

• numpy
• pandas
• chromedriver_autoinstaller
• selenium
• bs4
• urllib
• Python-Io
• datetime
• FinanceDataReader
• pykrx
• Ta-lib
• typing
• xgboost
• statsmodels
• torch

License

This repository is licensed under Apache 2.0.

Structure

└─ stock_prediction
    └─ stock_prediction
        ├─ setup.py
        └─ src
            ├─ stock_prediction_program
                ├─ __init__.py
                ├─ program_func.py
                ├─ my_data
                │   └─ *.csv
                ├─ my_news_data
                │   └─ *.csv
                ├─ my_package
                │   ├─ metrics.py
                │   ├─ model.py
                │   ├─ preprocessing.py
                │   ├─ train.py
                │   └─ visualization.py
                └─ my_path
                    ├─ bilstm
                    │  └─ *.pt
                    ├─ vae
                    │  └─ *.pt
                    └─ xgb
                       └─ *.joblib   

Usage

Performance

Additional Development Plans

• Analysis result output
• Used the lasted trend model
• Trading bot

Developer

Github: paulms77