---
title: "White-Box Prediction of Process Performance Indicators via Flow Analysis"
type: source
tags: [ppm, white-box, interpretable, flow-analysis, remaining-time]
authors: [Verenich Ilya; Nguyen Hoang; La Rosa Marcello; Dumas Marlon]
year: 2017
venue: "ICSSP 2017 (International Conference on Software and Systems Process), Paris"
kind: paper
raw_path: "raw/Predictive process monitoring/White-box prediction of process performance indicators via flow analysis icssp2017whitebox.pdf"
created: 2026-04-13
updated: 2026-04-13
key_claims:
  - Existing PPM methods are black-box — they predict a scalar without decomposition into elementary components.
  - A white-box approach predicts per-activity performance then aggregates via flow analysis.
  - Demonstrated for remaining cycle-time prediction on four real-life logs.
  - Competitive accuracy with interpretability gains.
---

# Verenich, Nguyen, La Rosa, Dumas 2017 — White-Box PPM via Flow Analysis

Combines **per-activity prediction** with [[methods/flow-analysis|flow analysis]] to produce an **interpretable (white-box)** remaining-time prediction: the overall estimate decomposes into contributions from each activity.

## Contribution
- Estimate cycle time of each remaining activity (the elementary component).
- Aggregate via classical flow-analysis rules (sequential/XOR/AND) to produce the instance-level prediction.
- Compared against black-box baselines — accuracy competitive; explainability substantially better.

## Significance
Early **interpretability-focused PPM**. Foreshadows [[concepts/explainability-apm|APM explainability]] and subsequent XAI-for-process-mining work (Mehdiyev & Fettke 2021 cited in [[sources/2026-calvanese-agentic-bpm-manifesto]] ref [57]).

## Connections
**Concepts:** [[concepts/remaining-time-prediction]] · [[methods/flow-analysis]] · [[concepts/explainability-apm]]
**Authors:** [[entities/ilya-verenich]] · [[entities/marlon-dumas]] · [[entities/marcello-la-rosa]]