Support Vector Machines for Binary Classification in Trading

Summary

This article introduces support vector machines as supervised classifiers that map labeled input examples into a feature space and determine a boundary separating two categories. It uses a two-dimensional diagram explanation and a hypothetical animal-identification task to show how multiple features can be used to classify a new observation. The training examples need labels from both classes, and the selected inputs should contain useful, relatively consistent distinctions between them.

The article then describes testing an SVM against a rule-based definition of the hypothetical class and discusses how classifier inputs and outputs are arranged in the implementation. Inputs are numeric values in a flattened array, while labels are Boolean values; the dimensions and sample counts must agree. It also cautions that an imbalanced set of labels can produce a poor classifier despite apparently quick training. The example is educational rather than a demonstrated trading system: it does not establish market performance, and the animal criteria and observations are constructed for illustration. Applying the approach to trading would require carefully chosen features, representative labeled data, and out-of-sample evaluation.

Key ideas

• An SVM learns a boundary that separates labeled observations into two classes.
• Each observation can be represented by multiple numeric features.
• Training data needs examples from both classes, with enough variation to learn a useful distinction.
• Input dimensions, flattened data length, and label count must match the classifier interface.
• The illustrative classification example does not demonstrate trading performance.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.