#### DMCA

## Enhancing Precision in Process Conformance: Stability, Confidence and Severity

Venue: | IEEE Symposium on Computational Intelligence and Data Mining (CIDM 2011 |

Citations: | 14 - 4 self |

### Citations

2069 |
Petri Nets: Properties, Analysis and Applications
- Murata
- 1989
(Show Context)
Citation Context ...can be represented with a formal model M . Given a model M and a trace σ, avail(σ,M) is the set of tasks available according to M after executing σ. Models for workflow representation like Petri nets =-=[13]-=- or its extensions (e.g., YAWL [14]) are considered in this paper. Due to the lack of space, the reader can follow the provided references to get insight into the theory underlying these models. An ex... |

487 |
Statistics for Experimenters. An Introduction to Design, Data Analysis, and Model Building.
- Box, Hunter, et al.
- 1978
(Show Context)
Citation Context ...fine the probability that a new trace in s′ follows the state s, and let 1−ps be the probability that the trace follows one of the other successor states of s′. According to the Binomial distribution =-=[21]-=-, the stability factor can be expressed as: Sft(s, τ) = l−1∑ i=0 ( z i ) (ps) i(1− ps)z−i (11) Formula (11) can be understood as follows: in order to s to still be an escaping state, i successes (ps)i... |

289 |
Workflow Mining: A Survey of Issues and Approaches
- Aalst, Dongen, et al.
(Show Context)
Citation Context ...blem of confronting the reality (in the form of logs) with the specification (in the form of formal models). Process mining comprises three classes of techniques: discovery, conformance and extension =-=[1]-=-. It can be oriented to the control flow, the data or the social aspects of a process. This work proposes process conformance techniques to aid the analysis of the control flow aspect. Process Conform... |

168 |
Theory of Linear and
- Schrijver
- 1986
(Show Context)
Citation Context ...e been defined, the maximum gain obtained with k traces must be computed. This problem is analogous to the well known Knapsack problem [18], which can be solved using binary integer programming (BIP) =-=[19]-=-. The following BIP model encodes the maximum gain over k traces: 1) Variables: The variable Xi ∈ {0, 1} (2) denotes if the escaping state i is covered or not. 2) Constraints: the total cost cannot ex... |

142 |
Finite Transition Systems
- Arnold
- 1994
(Show Context)
Citation Context ... example of a Petri net is shown in Fig. 1b. Finally we present transition systems, the formal models used in the development of the algorithms proposed in this paper. Definition 3 (Transition system =-=[15]-=-): A transition system (or TS) is a tuple (S, T,A, sin), where S is a set of states, T is an alphabet of actions, A ⊆ S × T × S is a set of (labeled) transitions, and sin ∈ S is the initial state. III... |

124 |
ter Hofstede. YAWL: yet another workflow language
- Aalst, M
- 2005
(Show Context)
Citation Context ...del M . Given a model M and a trace σ, avail(σ,M) is the set of tasks available according to M after executing σ. Models for workflow representation like Petri nets [13] or its extensions (e.g., YAWL =-=[14]-=-) are considered in this paper. Due to the lack of space, the reader can follow the provided references to get insight into the theory underlying these models. An example of a Petri net is shown in Fi... |

65 |
der Aalst. Conformance checking of processes based on monitoring real behavior
- Rozinat, van
(Show Context)
Citation Context ...nsion concerns about. Furthermore, in order to obtain a complete vision of the conformance, the precision measure should be included together with measures of the other dimensions like fitness (e.g., =-=[17]-=-). A consideration must be made concerning about the determinism of replaying a trace in the model. The approach presented is able to deal with duplicate tasks (i.e., several tasks in the model associ... |

59 |
Discovering Expressive Process Models by Clustering Log Traces
- Greco, Guzzo, et al.
(Show Context)
Citation Context ...e been presented in the literature, getting more robust-aware all the time (e.g. [3]). A complete survey can be found in [4], [5]. In particular, some examples of approaches focused on precision are: =-=[6]-=- (measuring the percentage of potential traces in the model that are in the log), [7] (comparing two models and a log to see how much of the first model’s behavior is covered by the second) used in [8... |

53 | Genetic Process Mining
- Aalst, Medeiros, et al.
- 2005
(Show Context)
Citation Context ...6] (measuring the percentage of potential traces in the model that are in the log), [7] (comparing two models and a log to see how much of the first model’s behavior is covered by the second) used in =-=[8]-=-, [9] (comparing the behavioral similarity of two models without a log), and [10] (using minimal description length to evaluate the quality of the model). The work presented in this paper extends and ... |

52 |
Process mining: a twostep approach to balance between underfitting and overfitting
- Aalst, Rubin, et al.
- 2010
(Show Context)
Citation Context ...0 0 3200 1435 1435 1435 946946947 764 764 764 54 54 54 H H G G G 0 G 1765 Fig. 2. Prefix Automaton 1The prefix automaton matches up with the transition system obtained using the approach presented in =-=[16]-=- with the parameters past, sequence and no horizon. B. Log-based exploration of Model Behavior The second step in the precision analysis is to determine the behavior of the model in order to compare i... |

35 |
A.: Process Discovery using Integer Linear Programming
- Werf, Dongen, et al.
- 2010
(Show Context)
Citation Context ...om [22]). These two benchmarks are based on the logs a32f0n00 5 and t32f0n00 5, both publicly available in [23]. The Petri net models used are the ones obtained using the Petri net discovery ILPMiner =-=[24]-=- ProM plug-in. The experiments conducted in Table I (b) focus on illustrating how the growth of a log influences the metric and its confidence given a particular selection of the stability and confide... |

23 | A.: Quantifying process equivalence based on observed behavior
- Medeiros, Aalst, et al.
- 2008
(Show Context)
Citation Context .... A complete survey can be found in [4], [5]. In particular, some examples of approaches focused on precision are: [6] (measuring the percentage of potential traces in the model that are in the log), =-=[7]-=- (comparing two models and a log to see how much of the first model’s behavior is covered by the second) used in [8], [9] (comparing the behavioral similarity of two models without a log), and [10] (u... |

16 | A.: Using Minimum Description Length for Process Mining
- Calders, Guenther, et al.
- 2009
(Show Context)
Citation Context ...g), [7] (comparing two models and a log to see how much of the first model’s behavior is covered by the second) used in [8], [9] (comparing the behavioral similarity of two models without a log), and =-=[10]-=- (using minimal description length to evaluate the quality of the model). The work presented in this paper extends and complements the technique introduced in [11], where the precision between a log a... |

16 | A Fresh Look at Precision in Process Conformance
- Munoz-Gama, Carmona
- 2010
(Show Context)
Citation Context ...ty of two models without a log), and [10] (using minimal description length to evaluate the quality of the model). The work presented in this paper extends and complements the technique introduced in =-=[11]-=-, where the precision between a log and a Petri net was considered. Using a log-guided traversal of the model, the refereed technique avoids the full state-space exploration of the Petri net, and prov... |

10 |
Leiserson, Introduction to Algorithms. McGraw-Hill Higher Education
- Cormen, Stein, et al.
- 2001
(Show Context)
Citation Context ... s#. Once the gain and the cost of covering an escaping state have been defined, the maximum gain obtained with k traces must be computed. This problem is analogous to the well known Knapsack problem =-=[18]-=-, which can be solved using binary integer programming (BIP) [19]. The following BIP model encodes the maximum gain over k traces: 1) Variables: The variable Xi ∈ {0, 1} (2) denotes if the escaping st... |

8 |
Robustness of community structure in networks, Phys
- Karrer, Levina, et al.
- 2007
(Show Context)
Citation Context ...= 57. The stability factor of the 2The idea of introducing perturbations in order to estimate some property has been used successfully in other fields, such as the measurement of community robustness =-=[20]-=-. imprecision at s is the probability of s still being an escaping state after considering z new traces, i.e., Sft(s, τ) = P z s (< l) = l−1∑ i=0 P zs (= i) (10) where P zs (< x) and P z s (= x) repre... |

6 |
der Aalst. Structural patterns for soundness of business process models
- Dongen, Mendling, et al.
- 2006
(Show Context)
Citation Context ...easuring the percentage of potential traces in the model that are in the log), [7] (comparing two models and a log to see how much of the first model’s behavior is covered by the second) used in [8], =-=[9]-=- (comparing the behavioral similarity of two models without a log), and [10] (using minimal description length to evaluate the quality of the model). The work presented in this paper extends and compl... |

5 |
der Aalst, “Towards an evaluation framework for process mining algorithms,” BPM Center Report BPM-07-06, BPMcenter.org
- Rozinat, Medeiros, et al.
- 2007
(Show Context)
Citation Context ...thin the precision dimension. Different algorithms for conformance checking have been presented in the literature, getting more robust-aware all the time (e.g. [3]). A complete survey can be found in =-=[4]-=-, [5]. In particular, some examples of approaches focused on precision are: [6] (measuring the percentage of potential traces in the model that are in the log), [7] (comparing two models and a log to ... |

3 |
A critical evaluation study of model-log metrics
- Weerdt, Backer, et al.
- 2010
(Show Context)
Citation Context ...the precision dimension. Different algorithms for conformance checking have been presented in the literature, getting more robust-aware all the time (e.g. [3]). A complete survey can be found in [4], =-=[5]-=-. In particular, some examples of approaches focused on precision are: [6] (measuring the percentage of potential traces in the model that are in the log), [7] (comparing two models and a log to see h... |

1 | Process Mining: Conformance and Extension - Rozinnat - 2010 |

1 |
PLG: a Framework for the Generation
- Burattin, Sperduti
- 2010
(Show Context)
Citation Context ... complex processes involving a large number of tasks and a high degree of parallelism, and some including also noise, duplicate and invisible tasks. These benchmarks have been generated with PLG tool =-=[25]-=-. In this table, the time reported corresponds to the time of computing only the metric, and the time also including the computation of the severity. The table illustrates that, even for large and com... |