This repository serves as a base model for predicting the closing price of a publicly traded equity. The first step of every model you create should start with a simple model then expand based off that. Every quantitative model should have different inputs based on the correlations to it.
For example:
- John Deere stock would correlate to commodity prices (wheat, corn, soybeans, etc.)
- Bank stocks would correlate with their respective countries' economy, interest rates, employment rate, etc.
This program was created through the use of XGBoost (a known algorithm for its speed and performance). For those who don't know, XGBoost is an algorithm, which continually places at the top of Kaggle (data science competitions). Rather than use a neural network to start off a new quantitative model, XGBoost is used to look at new variables first because of the speed.
- A number of inputs are put into the model with the target variable being the closing price
- As the inputs include variables such as the high and low of daily stock prices, these inputs would not be known until the end of the day so I set the target variable to be seven days into the future
- Note: Macroeconomic indicators were not included in this base model so the code can be adapted to predict commodity or index prices.
- MACD: Moving Average Convergence/Divergence Oscillator, a momentum indicator
- RSI: Oscillators help identify overbought and oversold markets
- ADX: Determine whether a market is in a trending or trading phase. It measures the degree of trend or direction in the market
- Average True Range: Measure of volatility
- Exponential Moving Average
- It should be noted that stocks with a higher beta are more difficult to predict based on the volatility
