---
title: "A Markov Prediction Model for Data-Driven Semi-Structured Business Processes"
type: source
tags: [ppm, markov, case-management, semi-structured, probabilistic]
authors: [Lakshmanan Geetika T.; Shamsi Davood; Doganata Yurdaer N.; Unuvar Merve; Khalaf Rania]
year: 2013
venue: "Knowledge and Information Systems (Springer)"
kind: paper
raw_path: "raw/Predictive process monitoring/A markov prediction model for data-driven lakshmanan2013.pdf"
created: 2026-04-13
updated: 2026-04-13
key_claims:
  - Semi-structured case-oriented processes require an instance-specific probabilistic process model, not a single global model.
  - Instance-specific PPMs can be transformed into Markov chains under non-restrictive assumptions.
  - An extended-space Markov chain handles parallel execution.
  - Prediction accuracy improves as more document data becomes available for the case.
---

# Lakshmanan et al. 2013 — Instance-Specific Markov PPM for Semi-Structured Processes

Early classical-probabilistic PPM. Targets **case-oriented, semi-structured** processes (insurance claims, healthcare, prescription orders) where activity order is determined by case-worker judgement, not a fixed workflow.

## Contribution
- Introduces an **instance-specific PPM**: transition probabilities are customised per case based on its document content.
- Shows how instance-specific PPMs can be lifted to a Markov chain (with an extended-space construction to handle parallel tasks).
- Demonstrates on a simulated automobile-insurance-claims process.

## Significance in the PPM landscape
Pre-deep-learning; representative of the classical probabilistic-model lineage that Evermann 2017 and Tax 2017 later contrasted their neural approaches against. Closely related to CMMN-style case management ([[frameworks/cmmn]]).

## Connections
**Concepts:** [[concepts/predictive-process-monitoring]] · [[concepts/outcome-prediction]] · [[concepts/next-activity-prediction]]
**Frameworks:** [[frameworks/cmmn]]
**Authors:** [[entities/geetika-lakshmanan]] · [[entities/merve-unuvar]]