Forest Proximities for Time Series

TL;DR

This paper introduces PF-GAP, an extension of RF-GAP, to proximity forests for time series classification, using forest proximities and MDS for embeddings, and exploring outlier detection connections.

Contribution

PF-GAP extends RF-GAP to proximity forests, improving time series classification and outlier detection through novel use of forest proximities.

Findings

Forest proximities better connect misclassified points and outliers.

Embeddings via forest proximities outperform traditional distance measures.

Proximity forests are efficient for time series classification.

Abstract

RF-GAP has recently been introduced as an improved random forest proximity measure. In this paper, we present PF-GAP, an extension of RF-GAP proximities to proximity forests, an accurate and efficient time series classification model. We use the forest proximities in connection with Multi-Dimensional Scaling to obtain vector embeddings of univariate time series, comparing the embeddings to those obtained using various time series distance measures. We also use the forest proximities alongside Local Outlier Factors to investigate the connection between misclassified points and outliers, comparing with nearest neighbor classifiers which use time series distance measures. We show that the forest proximities seem to exhibit a stronger connection between misclassified points and outliers than nearest neighbor classifiers.