---
title: "Intra and Inter-case Features in Predictive Process Monitoring: A Tale of Two Dimensions"
type: source
tags: [ppm, inter-case, remaining-time, feature-engineering, foundational]
authors: [Senderovich Arik; Di Francescomarino Chiara; Ghidini Chiara; Jorbina Kerwin; Maggi Fabrizio Maria]
year: 2017
venue: "BPM 2017, LNCS 10445, Springer: 306–323"
kind: paper
raw_path: "raw/Predictive process monitoring/Intra-and intercase senderovich2017.pdf"
created: 2026-04-13
updated: 2026-04-13
key_claims:
  - Prior PPM work assumes intra-case independence; this assumption breaks down when cases share scarce resources.
  - A bi-dimensional feature encoding captures both intra-case AND inter-case dependencies.
  - Case-type partitioning avoids feature-space explosion when encoding inter-case signals.
  - Evaluated on an Israeli ED log and a manufacturing log — remaining-time error improvements of 27% (ED) and 52% (manufacturing) respectively.
---

# Senderovich et al. 2017 — Intra- and Inter-Case Features

Foundational paper establishing **[[concepts/inter-case-ppm|inter-case features]]** as a first-class concern in [[concepts/predictive-process-monitoring|PPM]]. Targets [[concepts/remaining-time-prediction|remaining-time prediction]] specifically but the feature-encoding idea generalises.

## Contribution
- Bi-dimensional state-space feature encoding:
  - **Intra-case** — per-trace history (standard PPM features).
  - **Inter-case** — system state at the prefix cutoff: active cases, resource workload, queue lengths.
- **Case-type** partitioning — clusters of cases sharing the same state characteristics — keeps the feature space compact.
- Demonstrates large accuracy gains on two real-life logs.

## Significance
Directly cited as foundation for [[concepts/inter-case-ppm]]. Resurfaces in the APM Manifesto (queue-mining cited in passing) and in subsequent neural PPM work that bolts inter-case features onto LSTM/Transformer architectures.

## Connections
**Concepts:** [[concepts/inter-case-ppm]] · [[concepts/remaining-time-prediction]] · [[concepts/trace-encoding]]
**Authors:** [[entities/arik-senderovich]] · [[entities/chiara-di-francescomarino]] · [[entities/fabrizio-maggi]]