Preparing Financial Time Series Data for Deep Neural Networks
Summary
This article presents a data preparation workflow for deep neural network experiments on financial time series. It starts with terminal OHLCV quotes, organizes them by time, and derives price series and technical predictors, including adaptive trend filters, momentum measures, and oscillators. It also describes defining a target variable from an indicator or rule-based signal, then exploring distributions and preparing the dataset through cleaning, outlier analysis, and skewness assessment.
The author emphasizes that input preparation takes a substantial share of model development and strongly affects downstream results. The examples use R and a specific historical quote sample, with descriptive statistics offered to compare raw and outlier-adjusted data. The article focuses on data handling rather than demonstrating predictive trading performance, and its sample and selected predictors do not establish general suitability. The broader neural network series is described as using fully connected networks and planning later model training and evaluation work.
Key ideas
- The workflow begins with ordered OHLCV time series and constructs additional price-derived features.
- Technical filters, momentum measures, and oscillators are presented as candidate predictors.
- A target can be defined using an indicator or a sequence of conditional rules.
- Exploratory statistics and plots support data cleaning, outlier analysis, and skewness checks.
- Careful preparation is central to modeling, but this article does not establish predictive performance.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.