AI-driven process mining

Learn how to use AI-driven process mining in Exasol.

Process mining uses event log data to reconstruct and analyze business processes. By examining timestamps, activities, and case identifiers recorded in transactional systems, you can visualize how processes actually execute, identify bottlenecks, and find optimization opportunities.

Exasol’s AI Process Mining demonstrator shows how to run process mining workloads directly against an Exasol database. It combines Exasol’s fast analytical query engine with Python-based analysis to process large event logs without extracting data to a separate tool.

What process mining reveals

Business processes rarely execute the way they are designed. Orders get stuck in approval loops, support tickets bounce between departments, and manufacturing steps happen out of sequence. Process mining makes these patterns visible by reconstructing the actual process flow from event data.

Typical findings include:

  • Bottlenecks. Steps where cases spend disproportionate time waiting.
  • Rework loops. Activities that repeat because of errors or rejections.
  • Compliance deviations. Cases that skip required steps or follow unauthorized paths.
  • Throughput variations. Differences in processing time across regions, teams, or time periods.

How the demonstrator works

The AI Process Mining application is a Python-based tool that connects to an Exasol database containing event log data. It runs analytical queries against the live database and applies process mining algorithms to the results.

The demonstrator does the following:

  1. Reads event log data from Exasol tables (case ID, activity name, timestamp, and optional attributes).
  2. Reconstructs process flows by ordering events within each case.
  3. Identifies process variants (distinct paths through the process).
  4. Analyzes timing between activities to detect bottlenecks.
  5. Presents results through visualizations and summary statistics.

Because the queries run against the live Exasol database, the analysis reflects current data without requiring a separate data export step.

For details on visualization outputs and specific algorithms, see Process Mining with Exasol on GitHub.

Prerequisites

  • An Exasol database containing event log data

  • The AI Process Mining package installed

  • Python 3

For information about the supported Python version and additional system dependencies, see Process Mining with Exasol on GitHub.

Event log data format

Process mining requires event log data with at least three columns:

Column Description
Case ID Unique identifier for each process instance. For example: order number, ticket ID.
Activity The name of the process step. For example: Order Created, Payment Received
Timestamp When the activity occurred.

Additional columns (resource, department, cost) provide richer analysis but are not required.

A typical event log table in Exasol:

Copy 
CREATE TABLE process_events (
    case_id VARCHAR(50),
    activity VARCHAR(200),
    event_timestamp TIMESTAMP,
    resource VARCHAR(100),
    department VARCHAR(100)
);

Getting started

Install the AI Process Mining demonstrator from the Process Mining with Exasol GitHub repository:

Copy 
pip install git+https://github.com/exasol-labs/exasol-labs-ai-process-mining.git

For information about alternative installation methods and package names, see the repository READ ME.

After installation, configure the connection to your Exasol database and point the tool at your event log table. The repository README documents the configuration options and run commands for connecting to your database and specifying the event log source. The demonstrator handles the analysis pipeline from there.

Why Exasol for process mining

Process mining workloads are analytically intensive. Reconstructing process flows requires sorting millions of events by case and timestamp, computing time differences between consecutive activities, and aggregating results across process variants. These are operations where Exasol’s columnar storage and parallel query execution perform well.

Running the analysis against a live Exasol database has practical advantages:

  • No data movement. Event logs stay in the database. You do not need to export CSV files or load data into a separate process mining tool.
  • Fresh data. The analysis reflects the most recent events, not a periodic snapshot.
  • Scale. Exasol handles event logs with millions or billions of rows without requiring the data to fit in memory on a single analysis machine.

Use cases

Order-to-cash analysis. Trace orders from creation through fulfillment and payment. Identify where orders stall and which paths lead to the fastest completion.

IT service management. Analyze support ticket lifecycles to find resolution bottlenecks and escalation patterns.

Manufacturing. Track production steps to detect quality issues, rework patterns, and throughput constraints.

Compliance auditing. Verify that regulated processes follow required steps in the correct order.

Related resources