---
title: "Data-Aware Remaining Time Prediction of Business Process Instances"
type: source
tags: [ppm, remaining-time, data-aware, regression, pre-deep-learning]
authors: [Polato Mirko; Sperduti Alessandro; Burattin Andrea; de Leoni Massimiliano]
year: 2014
venue: "IJCNN 2014 (IEEE International Joint Conference on Neural Networks)"
kind: paper
raw_path: "raw/Predictive process monitoring/Data-Aware Remaining Time Prediction of IJCNN14.pdf"
created: 2026-04-13
updated: 2026-04-13
key_claims:
  - Remaining-time prediction benefits from jointly exploiting control flow AND data attributes of the running case.
  - Augments a transition system with time + data information at each state.
  - Prediction uses two factors - likelihood of subsequent activities (process model) + regression-based time estimation (data model).
---

# Polato, Sperduti, Burattin, de Leoni 2014 — Data-Aware Remaining Time Prediction

An early explicit **data-aware** remaining-time model, augmenting van der Aalst-style annotated transition systems with data-attribute-conditioned regression. Pre-deep-learning.

## Contribution
- Builds a transition system from historical log and annotates it with:
  - Activity-transition probabilities (control-flow perspective)
  - Regression-model time estimates conditioned on collected data (data perspective)
- Remaining time = likelihood-weighted sum over future paths.

## Significance
A cited baseline in subsequent PPM work including Polato et al.'s own later publications and the [[sources/2017-tax-lstm-process-prediction|Tax et al. 2017]] LSTM comparison. Establishes that **data features matter** — a conclusion confirmed by every subsequent generation of PPM models.

## Connections
**Concepts:** [[concepts/remaining-time-prediction]] · [[concepts/trace-encoding]] · [[concepts/predictive-process-monitoring]]