Deep Learning for Finance

Chapter 7. Machine Learning Models for Time Series Prediction

Machine learning is a subfield of AI that focuses on the development of algorithms and models that enable computers to learn and make predictions or decisions without being explicitly programmed, hence the term learning. Machine learning deals with the design and construction of systems that can automatically learn and improve from experience, typically by analyzing and extracting patterns from large amounts of data.

This chapter presents the framework of using machine learning models for time series prediction and discusses a selection of known algorithms.

The Framework

The framework is very important, as it organizes the way the whole research process is done (from data collection to performance evaluation). Having a proper framework ensures harmony across the backtests, which allows for proper comparison among different machine learning models. The framework may follow these chronological steps:

1. Import and preprocess the historical data, which must contain a sufficient number of values to ensure a decent backtest and evaluation.
2. Perform a train-test split, which splits the data into two parts where the first part of the data (e.g., from 2000 to 2020) is reserved for training the algorithm so that it understands the mathematical formula to predict the future values, and the second part of the data (e.g., from 2020 to 2023) is reserved for testing the algorithm’s performance on data that it has never seen before. ...