Research map / Learning theory and algorithms / Prediction and decision losses
Random Forests for Change Point Detection
Malte Londschien, Peter Bühlmann, Solt Kovács
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use a random forest classifier instead of a k-nearest-neighbor classifier in the classifier log-likelihood ratio (ablation) | changeforest vs changekNN · Simulated change points: Dirichlet | average adjusted Rand index: 0.69 -> 0.99 | Improved |
| use a random forest classifier instead of a k-nearest-neighbor classifier (ablation) | changeforest vs changekNN · Simulated change points: change in covariance | average adjusted Rand index: 0.01 -> 0.93 | Improved |
| use a random forest classifier instead of a k-nearest-neighbor classifier (ablation) | changeforest vs changekNN · Simulated change points: average over 9 setups | average adjusted Rand index: 0.81 -> 0.97 | Improved |
| increase trees from 100 to 500 and split trees to purity (ablation) | changeforest · Hyperparameter study | average adjusted Rand index and cost: computational cost fivefold | No clear effect |
| use extreme cheap settings (20 trees, depth 2, mtry 1) instead of defaults (ablation) | changeforest · Hyperparameter study (iris, breast cancer, wine) | average adjusted Rand index | Mixed |
| normalize covariates by the median absolute deviation of absolute consecutive differences | ECP, KCP, MNWBS · Nonparametric simulation setups | performance | Improved |
| use changeforest instead of ECP for dry beans | change point detection · Computation time: dry beans | average computation time (s): 5356 -> 2.6 | Improved |
| use changeforest instead of KCP for change in covariance | change point detection · Simulated change points: change in covariance | average adjusted Rand index: 0.95 -> 0.93 | Hurt |
| use changeforest model selection instead of heuristic or asymptotic selection (KCP, MNWBS, MultiRank) | change point detection · False positives on homogeneous data | percent with at least one change point: worst 5.00 vs 11.32 (KCP), 98.20 (MNWBS), 100.0 (MultiRank) | Improved |
| use changekNN model selection (same as changeforest) on breast cancer | changekNN · False positives on breast cancer | percent with at least one change point: 3.80 -> 29.80 | Hurt |
Closest papers by meaning
- Optimal Online Change Detection via Random Fourier Features (NeurIPS 2025 · similarity 0.86)
- Bayesian Model Selection for Change Point Detection and Clustering (ICML 2018 · similarity 0.86)
- Model-free Change-Point Detection Using AUC of a Classifier (JMLR 2025 · similarity 0.85)
- Optimistic Search: Change Point Estimation for Large-scale Data via Adaptive Logarithmic Queries (JMLR 2024 · similarity 0.85)
- Partially Observable Multi-Sensor Sequential Change Detection: A Combinatorial Multi-Armed Bandit Approach (AAAI 2019 · similarity 0.85)
- Uncertainty Quantification for Data-Driven Change-Point Learning via Cross-Validation (AAAI 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
