[Paper Review] From Internet of Things Data to Business Processes: Challenges and a Framework

Juergen Mangler, Ronny Seiger+13 more|arXiv (Cornell University)|May 14, 2024

Big Data and Business Intelligence4 citations

TL;DR

This paper proposes a generic, semi-automated framework to transform low-level Internet of Things (IoT) sensor data into higher-level business process events suitable for process mining. By structuring data abstraction into eight standardized steps—ranging from data discretization to expert-guided event correlation—the framework enables the derivation of process-aware event logs from raw IoT streams, with validation in a smart manufacturing use case demonstrating its ability to produce models comparable to expert-constructed ones.

ABSTRACT

The IoT and Business Process Management (BPM) communities co-exist in many shared application domains, such as manufacturing and healthcare. The IoT community has a strong focus on hardware, connectivity and data; the BPM community focuses mainly on finding, controlling, and enhancing the structured interactions among the IoT devices in processes. While the field of Process Mining deals with the extraction of process models and process analytics from process event logs, the data produced by IoT sensors often is at a lower granularity than these process-level events. The fundamental questions about extracting and abstracting process-related data from streams of IoT sensor values are: (1) Which sensor values can be clustered together as part of process events?, (2) Which sensor values signify the start and end of such events?, (3) Which sensor values are related but not essential? This work proposes a framework to semi-automatically perform a set of structured steps to convert low-level IoT sensor data into higher-level process events that are suitable for process mining. The framework is meant to provide a generic sequence of abstract steps to guide the event extraction, abstraction, and correlation, with variation points for plugging in specific analysis techniques and algorithms for each step. To assess the completeness of the framework, we present a set of challenges, how they can be tackled through the framework, and an example on how to instantiate the framework in a real-world demonstration from the field of smart manufacturing. Based on this framework, future research can be conducted in a structured manner through refining and improving individual steps.

Motivation & Objective

• Address the challenge of extracting meaningful process events from low-granularity IoT sensor data in industrial and service environments.
• Bridge the gap between raw IoT data and process-aware analytics by enabling the semi-automated derivation of process events from sensor streams.
• Support domain experts in creating process models from IoT data without requiring deep expertise in data engineering or process mining.
• Provide a structured, extensible framework to guide the transformation of raw sensor data into enriched event logs for process mining.
• Demonstrate the framework’s feasibility and quality through a real-world smart manufacturing case study, showing alignment with expert-constructed process models.

Proposed method

• Define a generic 8-step framework to convert raw IoT sensor data into process-aware event logs, with abstraction and correlation steps for higher-level event modeling.
• Apply data discretization (Step 1) to convert continuous sensor values into discrete events, using time-based or value-based thresholds.
• Identify sensor groups (Step 2) based on temporal proximity and functional relevance to form potential process event clusters.
• Construct topologies (Step 5) to model relationships between sensor events, using time windows and dependency analysis to infer event sequences.
• Use statistical and machine learning techniques (e.g., correlation, clustering) in Steps 3–5 to detect patterns and dependencies among sensor events.
• Incorporate expert input in Steps 6 and 7 to refine event boundaries, groupings, and abstractions, ensuring domain relevance and model quality.

Experimental results

Research questions

• RQ1 How can low-level IoT sensor data be systematically abstracted into higher-level process events suitable for process mining?
• RQ2 What are the key challenges in mapping raw sensor streams to process-level events, and how can they be addressed through a structured framework?
• RQ3 To what extent can a semi-automated framework produce process models of quality comparable to those created by domain experts?
• RQ4 How can the framework be instantiated and validated in real-world industrial scenarios such as smart manufacturing?
• RQ5 What role do domain experts play in refining the output of automated data abstraction steps, and how can their input be effectively integrated?

Key findings

• The framework enables the transformation of raw IoT sensor data into process-aware event logs through a sequence of eight structured, extensible steps.
• The approach supports semi-automated derivation of process events, reducing reliance on data engineers and allowing domain experts to guide abstraction and correlation.
• In a smart manufacturing case study, the framework-generated process model achieved comparable quality to expert-constructed models, validating its practical utility.
• Performance bottlenecks were identified primarily in Step 1 (discretization) and Step 4 (dependency calculation), especially with high-volume sensor data.
• The framework is adaptable to other domains such as healthcare, as it relies only on time-series sensor data and does not depend on domain-specific logic.
• User interfaces for expert interaction in Steps 6 and 7 are critical for maximizing framework utility and model accuracy.