Process discovery: Inductive Miner Sander Leemans D. Fahland W.M.P van der Aalst
Process discovery system log process model S.J.J. Leemans
Quality = Sound Simple fast Fitting Precise System behaviour Complete log Recorded log fast Fitting Precise = Log Discover model S.J.J. Leemans
Results with incomplete logs not fitting not sound not simple not sound Flower model not precise α ILP Heuristics Miner Evolutionary Tree Miner not fitting not sound not fast
Outline {<a,b,c>, <a,c,b>, <a,d,e>, System behaviour {<a,b,c>, <a,c,b>, <a,d,e>, <a,d,e,f,d,e>} = S.J.J. Leemans
Block-structured Petri nets τ a b c d e τ x /\ Activities Sequence Loop Parallel Exclusive choice Sound a b c d e → Sander Leemans
Outline System behaviour = S.J.J. Leemans
Divide & conquer a {<a,b,c>, <a,c,b>, <a,d,e>, <a,d,e,f,d,e>} {<a>, <a>, <a>} {<b,c>, <c,b>, <d,e>, <d,e,f,d,e>} recurse recurse S.J.J. Leemans
Finding operator {<b,c>, <c,b>, <d,e>, <d,e,f,d,e>} recurse a b c d e f a {<a,b,c>, <a,c,b>, <a,d,e>, <a,d,e,f,d,e>} Find cut in directly-follows graph Sequence: edges crossing one-way only S.J.J. Leemans
recurse … {d,e,f} {b,c} { , } { <b,c> <c,b> , <d,e> a x {<b,c>, <c,b>, <d,e>, <d,e,f,d,e>} { , } { <b,c> <c,b> , d e b c f {<b,c>, <c,b>} {<d,e>, <d,e,f,d,e>} <d,e> <d,e,f,d,e> { , } } Exclusive choice: no crossing edges S.J.J. Leemans
… one more recursion … {f} {d,e} {< < >, >} {< d,e x a {< < >, >} {< d,e >, {<b,c>, <c,b>} {<d,e>, <d,e,f,d,e>} < d,e , f , {< >} d,e >} {<d,e>} {<f>} f e d d e f Loop: identify body and loopback parts (assumption: start/end activities disjoint) S.J.J. Leemans
… last recursion {b} {c} {< < > >} {< < b , c >, x f a e d {b} {c} {< < > >} {< < b , c >, >} {< < > >} {<b,c>, <c,b>} c b {<b>} b {<c>} c b c Parallel: all possible crossing edges S.J.J. Leemans
Result τ a b c d e f x f a e d b c S.J.J. Leemans
No cut y z x a x x y z b c S.J.J. Leemans
? Inductive Miner ? Divide activities Select operator Split log Else: flower model Split log Recurse on splitted logs {c,d} {a,b} ? {c} {d} Sander Leemans
Outline System behaviour = S.J.J. Leemans
Rediscoverability Complete log System behaviour = (language equivalent) = (normal form) Log Discover model Block-structured with Start\end activities of loop disjoint No duplicate activities No silent activities (τ) x Directly-follows graph complete Noise-free S.J.J. Leemans
Incomplete logs Sound Simple polynomial Fitting Most precise System behaviour Complete log Sound Simple Incomplete log polynomial Fitting Most precise (by framework; bring your own operator) S.J.J. Leemans
Future Work Generalise Block-structured Start\end activities of loop must be disjoint No duplicate activities No silent activities (τ) Directly-follows graph complete Noise-free x S.J.J. Leemans
You have been watching Sound Simple polynomial Fitting Most precise System behaviour Complete log Sound Simple Incomplete log polynomial Fitting Most precise (by framework; bring your own operator) S.J.J. Leemans