Download presentation
Presentation is loading. Please wait.
1
ENHANCING PRECISION IN PROCESS CONFORMANCE Stability, Confidence And Severity
JORGE MUNOZ-GAMA and JOSEP CARMONA Universitat Politecnica de Catalunya Barcelona, Spain
2
Conformance: precision
Process Information System Logs Model Discovery ? Conformance Generalization Structure Fitness Precision
3
(1) Log Behavior Prefix automaton of log behavior 1435 A B D E A 946
# Instances Log Traces H D F A 54 1435 A B D E A 946 A C D G H F A 764 A C G D H F A G D H F A 764 54 A C G H D F A 1 A C D G G H F A 818 764 A C D G H F A 946 2381 3200 3145 1435 3199 3199 3145 3200 2381 1435 1765 1764 1710 946 947 946 947 946 A F H G 1 B D E A 1435 1435 1435 1435
4
(2) Log-based Model Exploration
Extend with tasks availed by the model in each state B E A E B D C G H F 946 1 1435 764 54 818 947 3200 1765 A D A C H F G G G H H G
5
(3) Comparing Log and Model
Imprecisions = in the model but not in the log Threshold ( ) for robustness A E B D C G H F 946 1435 764 54 818 947 3200 1765 G G H H G H F A 1 G 1 1 1
6
* Extension of Munoz-Gama
Metric Counts and weights imprecisions according to their frequencies Estimating the effort needed to achieve a model completely precise A F H G 1 E B D C 946 1435 764 54 818 947 3200 1765 * Extension of Munoz-Gama and Carmona BPM 2010
7
Confidence log K Low Confidence High Confidence
Related with imcompletness and more traces more confidence Low Confidence High Confidence
8
Confidence: Upper Estimation
Best scenario = covering imprecisions D F A BIP Formulation K = 3 H 54 54 54 54 D H F A G G 818 G 764 764 764 764 A C D G H F A 3200 3200 1765 H 947 H 947 G 946 946 946 B 1 Upper Bound D E A 1435 1435 1435 1435 Cost of an imprecision (C): Gain of an imprecision (G):
9
Confidence: Lower Estimation
Worst scenario = new escaping states K = 1 G 1 A C D H F 946 764 54 818 947 3200 1765 1 Lower bound new states with escaping states each e.g.
10
Confidence Results
11
Severity All imprecisions equally important?
D F A D F A H 54 H 54 H 54 54 H 54 54 H 54 H 54 D F H H A H H D F A D H F A D H F A H 54 54 54 54 H 54 54 54 54 818 G G G 764 764 764 818 G G 764 764 764 764 764 D H F A D H F A 818 G G G G sever mid low A C D G H F A G 764 764 764 818 764 764 764 764 764 3200 3200 1765 H 947 H 947 G 946 946 946 H F A A C D G H F A B 1 G 1 1 H 3200 1 H 3200 1765 H 947 H 947 G 946 946 946 D E A H F A B 1435 H 1435 H 1435 H 1435 H 1 G 1 1 1 D E A H H 1435 1435 H 1435 H 1435 All imprecisions equally important? Subjective and multifactor Frequency, Alternation, Stability, Criticality
12
Severity: Frequency Imprecision in frequent parts more sever sever
10000 7000 3000 10 7 3 sever sever
13
Severity: Alternation
More chances to make a mistake more sever sever sever
14
Severity: Stability Apply perturbation
increase the number of instances in that point proportional to the current occurrence number Measure how easy is to overpass the threshold Imprecision stable to perturbation more sever sever sever 3000 3000 10000 7000 10000 6901 99
15
Severity: Criticality
Importance of the task involved in the imprecision Inspired on Cost-based Fitness in Conformance Checking by Adriansyah, Sidorova and van Dongen, ACSD 2011 sever sever Bank Transfer Check Date Format
16
Severity Results * Benchmarks produced by PLG
by Andrea Burattin and Alessandro Sperduti
17
Implementation ETConformance Plug-in
18
Not addressed in this presentation
Non fitting traces Invisible and Duplicate tasks Conclusions Metric to measure the precision Confidence interval over the metric Severity assessments over the imprecisions Implemented in an open-source framework
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.