An Encoding Approach for Stable Change Point Detection

TL;DR

This paper introduces a nonparametric, encoding-based method for stable change point detection in multivariate time series, effectively estimating the number and locations of change points without prior distributional assumptions.

Contribution

It develops a novel encoding and aggregation framework combined with a searching algorithm for accurate change point detection in multivariate data.

Findings

Outperforms existing nonparametric methods in simulations

Effective on categorical, ordinal, and continuous data

Provides stable and consistent change point estimates

Abstract

Without imposing prior distributional knowledge underlying multivariate time series of interest, we propose a nonparametric change-point detection approach to estimate the number of change points and their locations along the temporal axis. We develop a structural subsampling procedure such that the observations are encoded into multiple sequences of Bernoulli variables. A maximum likelihood approach in conjunction with a newly developed searching algorithm is implemented to detect change points on each Bernoulli process separately. Then, aggregation statistics are proposed to collectively synthesize change-point results from all individual univariate time series into consistent and stable location estimations. We also study a weighting strategy to measure the degree of relevance for different subsampled groups. Simulation studies are conducted and shown that the proposed change-point methodology for multivariate time series has favorable performance comparing with currently popular nonparametric methods under various settings with different degrees of complexity. Real data analyses are finally performed on categorical, ordinal, and continuous time series taken from fields of genetics, climate, and finance.