alphaQ

A Q-learing based portfolio trading system built using OpenAI Gym and Stable-Baselines3.

Requirements

• Python 3.7

Installation

1. Clone this repository: git clone https://github.com/mannmann2/alphaQ.git
2. cd alphaQ
3. pip install -r requirements.txt

Usage

For learning how to train and evaluate the agents, follow the starter notebook provided with this repository.

• main.ipynb

Or view it in Jupyter's online notebook viewer:

• Nbviewer

Results

Models

Below is a list of all the pre-trained models provided.

Best models for [AAPL, JPM, MSFT, V]

DQN_best
DDPG_best

Alternate models for [AAPL, JPM, MSFT, V]

DQN: DQN 5, DQN 6, DQN 7, DQN 8
DDPG: DDPG 7, DDPG 8

Variant Models

DQN A AXP, CVX, DIS, KO
DQN B JNJ, MCD, MMM, WMT
DQN C CAT, CSCO, HD, IBM

DDPG A JNJ, MCD, MMM, WMT
DDPG B AMGN, NKE, UNH, VZ
DDPG C GS, NKE, PG, UNH

Documentation

Project Proposal
Progress Slides 1
Progress Slides 2
Progress Slides 3
Progress Slides 4
Preliminary Report
MSc Final Report

• Free software: MIT license