BOSS Symbolic Time-Series Classification Compared with Dynamic Time Warping
Summary
The article presents Bag-of-SFA-Symbols (BOSS) as a way to classify market regimes by converting price windows into symbolic words. Each window is z-normalized to reduce sensitivity to price level, transformed with a low-pass Fourier representation to retain broad shape, and quantized into letters. A sliding window turns a longer segment into a bag of words, which can be compared with other segments. The article describes an ensemble of BOSS classifiers and a rule-based labeling method for trending, ranging, and volatile regimes, then benchmarks the approach against Dynamic Time Warping (DTW).
The reported comparison is mixed: a single BOSS model can fall short, while the ensemble competes more effectively and offers interpretable vocabularies. The article also discusses sensitivity to noise and possible improvements such as richer labels and alternative weighting methods. Its findings are tied to the presented labeling scheme and benchmark data; regime labels are rule-generated rather than independently verified market states, and classification performance alone does not establish profitable trading utility.
Key ideas
- BOSS represents normalized price windows as words derived from low-frequency Fourier coefficients.
- A bag of symbolic words summarizes a segment without preserving each word’s exact position within it.
- The article compares single and ensemble BOSS classifiers with DTW for regime classification.
- The regime labels are generated by rules using autocorrelation and volatility rather than manual annotation.
- Results are mixed, and classification accuracy does not by itself demonstrate trading profitability.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.