---
title: "Cycle Time Prediction: When Will This Case Finally Be Finished?"
type: source
tags: [ppm, remaining-time, non-parametric-regression, foundational, pre-deep-learning]
authors: [van Dongen Boudewijn F.; Crooy R. A.; van der Aalst Wil M. P.]
year: 2008
venue: "OTM Conferences 2008, LNCS 5331, Springer"
kind: paper
raw_path: "raw/Predictive process monitoring/Cycle time prediction when will this 2008.pdf"
created: 2026-04-13
updated: 2026-04-13
key_claims:
  - Remaining cycle time for a running case can be predicted from historical logs using non-parametric regression.
  - Non-parametric regression handles the sparse-precedent setting typical of process data.
  - Average cycle time is a poor predictor for mid-process cases; historical attribute-conditioned prediction is substantially better.
---

# van Dongen, Crooy, van der Aalst 2008 — Cycle Time Prediction

One of the earliest papers on [[concepts/remaining-time-prediction|remaining cycle-time prediction]]. Pre-dates deep learning. Uses **non-parametric regression** over features extracted from the partial trace.

## Contribution
- Five predictor variants (different feature extraction strategies) are compared on a real-life administrative log.
- Establishes that the naïve baseline (average cycle time minus elapsed) is poor mid-process — a key motivating observation for subsequent PPM research.

## Significance
Cited by [[sources/2017-tax-lstm-process-prediction|Tax et al. 2017]] as the non-parametric-regression baseline. Represents the **pre-deep-learning lineage** of remaining-time work that [[entities/wil-van-der-aalst]] co-led.

## Connections
**Concepts:** [[concepts/remaining-time-prediction]] · [[concepts/predictive-process-monitoring]]
**Authors:** [[entities/wil-van-der-aalst]]