/490-poc

Playground for quant experiments, a proof of concept.

Primary LanguagePython

490-poc

Examining the feasibility of quant methods in finance (a proof of concept).

To run

Initialize a python virtual environment and pip install requirements (keras, transformers, numpy, pandas, csv).

  • LSTM - run python testing.py from the root directory
  • BERT - cd into sentiment/, then run python emotion.py

Models

Testing various RNN models to find strengths/weaknesses of each as it applies in micro & macroeconomic theory.

  • LSTM: univariate (vanilla, stacked, bidirectional); multivariate; multiparallel
  • NLP: BERT (Bidirectional Encoder Representations from Transformers)

Data

Data in csv format, preprocessed to calculate log returns.

  • S&P 500: weekly intervals for 20 years
  • VIX index to measure volatility