Hands-On AI Trading with Python, QuantConnect, and AWS

Chapter 3Step 1: Problem Definition

The problem definition step in algorithmic trading focuses on the identification of the specific financial objective the algorithm aims to achieve, such as predicting stock prices, optimizing trade execution, or managing risk through dynamic portfolio adjustments.

The chosen financial objective is then translated into the target variable you intend to predict or optimize—for example, a stock’s future price, the next period’s market volatility, or the expected return on a portfolio.

Once the target variable is identified, the next step is to specify the scope and constraints of the problem, such as defining the time frame for predictions (e.g., intraday, daily, weekly), the markets or assets to be included (e.g., equities, commodities, forex), any regulatory or operational constraints that must be considered, as well as the trading strategy’s risk tolerance and performance benchmarks.

A well-defined problem also includes understanding the relationships between the target variable, known as labels or dependent variables, and potential predictor variables, known as features or factors or independent variables. These features can include historical price data, trading volumes, economic indicators, and even sentiment analysis from news articles or social media. Establishing a clear hypothesis about how these features interact with the target variable will inform the data collection and feature engineering processes.

Let’s illustrate this step with ...