Applications of Machine Learning in Trading

This lecture will cover the essence of ML and its pivotal role in data analysis. We'll discuss various data types that fuel ML algorithms, delve into the core categories of ML (supervised, unsupervised, and reinforcement learning), and unravel the significance of correlation, regression, and cluster analysis in extracting meaningful insights from complex datasets.

The lecture will consider data cleaning, specifically the handling of missing values and the common functions to use in python for that. It will also examine Data Transformations such as common scaling methods. And finally will consider Data Reduction where we consider the tool of principal component analysis and how to visually understand the way in which it helps to reduce dimension.

This lecture examines signal levels and the fact that  "A strong signal emerges if weak signals are pointing in the same direction."

This lecture considers both classification as well as regression metrics.

This lecture examines time series analysis where we consider both stationarity and the importance when using ARIMA. It also examines how the random forest makes use of an ensemble approach to leverage the strength of each individual model to improve overall prediction accuracy and robustness against overfitting.

We examine the general problem statement for portfolio optimization along with noise-induced instability.

We investigate the Marcenko-Pastur probability density function as well as how it differs from that of the Ledoit-Wolf method.

We consider de-noising by identifying and adjusting noise-related eigenvalues.

This lecture examines Nested Cluster Optimization along with the difference between inter-clustering and intra-clustering.

Examine how to go about applying optimization methods including NCO.

After determining optimal allocations with NCO, Monte Carlo simulations can be used to test these allocations under varied market conditions.

We examine how to analyze the performance of NCO on different types of portfolios (like minimum variance and maximum Sharpe ratio

This lecture will examine trading volume and why it is valuable in predicting future price movements.

In this lecture we will consider the prediction of trading volume, especially the intraday change. This is a relatively underexplored area for financial forecasting.

In this lecture we will examine how to go about using meta-learning to capture both common patterns and our stock-specific patterns.

In addition, we will consider how to  employ an encoder-decoder framework to generate latent variables for better prediction.

In this lecture we examine the intersection of technology and human emotion. We will investigate how algorithms can understand not just words, but the meaning behind them as well.

This lecture examines sentiment analysis. It specifically analyses how to go about making use of news sentiment specifically when dealing with foreign exchange futures strategies.

This lecture examines a case study by Bloomberg news and social sentiment data.

Consider a few applications with satellite imagery as alternative data.

Predicting Stock Market returns with satellite imagery.

Predicting Stock Market returns with satellite imagery part 2.

Examine how knowledge-driven AI and data-driven AI can be combined in decision-making.

Investigate the traditional workflow.

Understand the 3 modules found in the automated pipeline.

Examine the factor mining pipeline.

Finding hyperparameters by use of search algorithms.

Understand the meaning of Individual Conditional Expectation (ICE) & Partial Dependence Plots (PDPs).

Examine LIME.

Examine model explainability.

This lecture examines the role of artificial intelligence in finance as well as alternative data.

This lecture examines how to apply ML and XAI (Explainable Artificial Intelligence) technologies to improve the financial performance of Statistical Arbitrage strategies.

This lecture will examine multi-period asset allocation during different investment periods. It considers whether asset allocation problems can be represented using RDDL   (Relational Dynamic Influence Diagram Language).

This lecture examines Natural Language Processing (NLP), a fusion of computer science and linguistics that utilizes algorithms to understand and interpret human language.

This lecture examines the structure of an earnings call. It will also examine the medium effect of big data. Note that page 46 of the paper can be considered for extra reading.

This lecture examines the application of Natural Language Processing (NLP) and machine learning in enhancing investment strategies by analyzing conference call transcripts.

It will investigate different NLP classifiers, including the Loughran-McDonald sentiment dictionary, the FinBERT model, and Alexandria Technology’s ML Ensemble.

This lecture compares the three NLP Classifier strategies. It also considers breaking down returns in order to identify how much of the strategy's performance is due to these known risk factors and how much is unique or "alpha."