stock_bot

A machine learning model trained on historical market data to trade using papermoney

Overview

• Data Ingestion: Fetches historical bars (price, volume) using Alpaca's market data endpoints.
• Signal Generation: Processes data through a simple MLP network to predict price movement directions.
• Order Execution: Places buy/sell orders automatically based on model confidence using the Alpaca Python SDK.

Dependencies

• Pytorch
• Pandas
• Numpy
• Alpaca API
• Version: Python 3.x

Prereqs

• An Alpaca Account (Paper trading is recommended for testing).

Installation

1. Clone Repo

  git clone https://github.com/funnymonke0/stock_bot

2. Install Dependencies

  cd simple-isef-scraper
  pip install -r requirements.txt

Model

The core of this bot is a Multi-Layer Perceptron. It is trained on historical features

• Log Volume (20 bar normalized volume)
• Momentum/return (log normalized return from the previous bar)
• Log OHLV (log normalized open, high, low, close based on previous close)
• I chose to keep these inputs simple normalized ohlcv values, but will update at a later date

##Results (confusion matrices, actual accuracies vary around 40%, which I have not graphed yet)

• Results tab
• Example for crypto model version 1.6:
•