FlowMine — process mining for Python

FlowMine turns an event log (one row per activity: case id, activity, timestamp) into a process map, performance metrics and AutoInsights: an automatic check of every activity for bottlenecks, rework, loops, one-off incidents, growing durations and rare steps, each with the time it costs and a money estimate at your hourly rate.

What sets it apart from general process-mining libraries:

• AutoInsights with an effect estimate — one explained finding per activity (AutoInsights.findings()), in Russian or English.
• Russian-first data handling — dd.mm.yyyy dates, cp1251 files, Russian log messages (switchable to English).
• Simulation and ML helpers — process simulation (flowmine.imitation), missing value imputation, NLP helpers for free-text columns (flowmine[nlp]).

For a web UI on top of this library, see FlowMine Studio.

Install

pip install flowmine              # core
pip install "flowmine[nlp]"       # + torch/transformers text features
pip install "flowmine[embeddings]"  # + graph-embedding loop detection (gensim, catboost)
pip install "flowmine[full]"      # everything

Graphviz drawing (GraphvizPainter) also needs the Graphviz executables on your PATH: https://graphviz.org/download/

Quick start

import pandas as pd
from flowmine import DataHolder
from flowmine.autoinsights import AutoInsights
from flowmine.miners import HeuMiner

df = pd.DataFrame({
    "case":     ["c1", "c1", "c1", "c2", "c2", "c2", "c2", "c2"],
    "activity": ["Submit", "Check", "Approve", "Submit", "Check", "Request info", "Check", "Approve"],
    "start":    pd.to_datetime([
        "2026-03-01 09:00", "2026-03-01 09:30", "2026-03-01 11:00",
        "2026-03-02 10:00", "2026-03-02 10:20", "2026-03-02 12:00",
        "2026-03-03 09:00", "2026-03-03 10:00",
    ]),
})

holder = DataHolder(df, col_case="case", col_stage="activity", col_start_time="start")

miner = HeuMiner(holder)
miner.apply()                      # miner.graph: the discovered process model

insights = AutoInsights(holder, min_cost=50 / 60)   # cost of one minute of work
insights.apply()
for finding in insights.findings(lang="en"):
    print(finding["stage"], [r["title"] for r in finding["reasons"]], finding["financial_effect"])
print(insights.fin_effects_summary(lang="en", currency="USD"))

Data can also be a path: .csv, .xlsx, .txt or .xes / .xes.gz (flowmine.read_xes() reads XES into a DataFrame).

What's inside

Language

Log messages are in Russian by default. For English:

export FLOWMINE_LANG=en        # or: flowmine.set_language("en")

AutoInsights.findings() and fin_effects_summary() take lang="ru" / lang="en".

Changes

See CHANGELOG.md.

License

MIT. © FlowMine Contributors.