A project trying to generate stock marker predictions using a combination of technical indicators and google trends to explain some of the up and downs. This project is made for educational purposes only and still needs lots of refinements over the predictor model.
To visualize the model, there is a frontend UI made with Electron and React. The prediction and data logic is done with python and tensorflow.
- Frontend
- Node v14.15.x
- Yarn v1.22.x.
- Install libraries using
yarn install
- Backend
- Python 3.7
- Tensorflow 2.0.0 or greater
- The following packages corresponding to python 3.7:
- sklearn
- numpy
- flask
- alpha_vantange (wrapper)
- pytrends
- pandas
NOTE: Everything except alpha_vantage is available through anaconda's conda-force channel. alpha_vantage is available through pip.
Right now the model goes through one input layer of google trends, one layer of the stock data and two layers, sma(simple moving average) and macd(moving average convergence divergence) to train the model.
The training phase will go through 5 epochs with batch size of 32.
yarn installyarn start- After backend server starts, go to settings and add alpha_vantage
API_KEY. Click here to get anAPI_KEY.
- Move
config_example.initoconfig.ini. - Activate virtual environment if any.
- Make sure to add
backenddirectory topython path. (IMPORTANT) python backend/server.py
The UI will need the backend to be running in order to fetch datasets and do predictions.
There are launch and task scripts made for vscode.
Start by running the server, then either Electron:Main to just run instance of Electron (can debug only main process) or Electron:all which can debug the frontend ui as well.
Coming soon.