2-AI Trading Strategies

Welcome to Udacity. Takes 5 minutes to get familiar with Udacity courses and gain some tips to succeed in courses.

You are starting a challenging but rewarding journey! Take 5 minutes to read how to get help with projects and content.

Learn how to prepare price data for AI models, backtest trading algorithms, and build a simple RSI algorithm.

Explore investment data, summarize key stats, use K-Means and PCA for clustering, adapt trading algorithms, and identify risk factors to enhance model insights on outperformance

Build regression models using past returns, explore regularization to avoid over/underfitting, and differentiate between training and test data while identifying signs of overfitting and underfitting.

Predict categorical variables using logistic regression and decision trees. Improve model performance with cross-validation for strong out-of-sample results.

Explore reinforcement learning (RL) and its components, Q-learning, the DQN algorithm, and how to build and backtest an RL model.

We'll define the steps of the machine learning pipeline, from data ingestion to production. We'll emphasize preprocessing and feature engineering, essential steps to well-performing trading models.

How do you get data from the Internet to your model? We'll talk about ingestion, transformation,and data wrangling using Pandas, the industry-standard time-series package for data manipulation.

Feature engineering significantly improves model performance, and in this lesson, we'll go over strategies you can use. Basic and more involved techniques will be presented, with use cases noted.

What's your data trying to tell you? Use EDA, exploratory data analysis, to find out! We'll use Python's two most popular plotting packages, matplotlib and Plotly, to find out your data's secrets.

Learners will use historical stock prices for two large companies to practice data manipulation and exploratory data analysis, or EDA.

Understand the foundations for backtesting. Learners will examine formulas and develop tools for calculating and plotting returns.

Understand volatility, skewness, kurtosis, and the impacts that these concepts have on developing a trading strategy.

Explore drawdowns, how to calculate them, and which ratios should be employed when developing a backtest strategy. Calculations include the Sharpe, Sortino, and Calmar Ratio.

Through Python, learn how to implement Walk-Forward Validation and combine core calculations to develop a robust backtesting strategy.

Assess and manage investment risk through key calculations such as Volatility, Sharpe Ratio, Sortino Ratio, Calmar Ratio. Learners develop and backtesting a strategy using Walk-Forward Validation.

Introduction to reinforcement learning, Q-learning, and core concepts including how reinforcement learning fits in the trading world.

Explore the concept of Financial State and Action Spaces. Learn how to define states and extract popular market indicators and conditions with Python and YFinance.

Construct a RL trading model using Python including define and running a training loop. Learn key tips for implementation and run test data through the newly created model.

Examine key backtesting concepts, gather important backtesting information on an RL model, and learn how to interpret those results to optimize performance.

The Project for this course will involve the students building and training RL Q-learning agent from scratch in a jupyter notebook.

We review how AI models work in principle and important terminology used in AI model training and optimization. We talk about where AI model optimization applies in using AI models for trading.

Overfitting is a common issue when training AI models for trading. We’ll explore bias, variance, and the role of hyperparameters in the context of various AI model types.

Get hands-on with AI model hyperparameters and discuss the various methods available to us for tuning them in a systematic or ad-hoc way, as well as the advantages and disadvantages of each method.

We discuss some practical methods and important considerations related to model optimization and evaluation in the context of AI models for trading.

We analyze important practical considerations for using AI models for trading. We discuss things we need to keep in mind as we maintain or iterate on our deployed models.

Optimize a stock price prediction model using data preprocessing, hyperparameter tuning, over/underfitting detection, model evaluation, and feature selection.

Normal distribution and geometric Brownian motion are key to momentum trading. Shapiro-Wilk’s and Student’s t-tests are useful statistical tools for developing momentum-based trading strategies.

Explore geometric Brownian motion for stock price modeling, calibration, forecasting, and confidence intervals, followed by deriving and coding the Black-Scholes formula for European option pricing.

Building a momentum-based trading program using MySQL/SQLite, Python, and geometric Brownian motion for price forecasting, confidence intervals, and Monte-Carlo simulation for scenario analysis.

The lesson covers back-testing momentum strategies, evaluating with Sharpe ratio and maximum drawdown, and quantitative risk management using Value-at-Risk (VaR) and Expected Shortfall (ES) in Python.

Build a momentum-based strategy to trade the S&P 500 index. You can later expand and customize you project to suit your needs. You will use the Python packages numpy, scipy and sqlite3, among others.

Congratulations on your graduation from this program! Please join us in celebrating your accomplishments.