Category: Example discovery algorithm

Details The Trace Miner discovers the base model where every unique trace has its own dedicated path through the model. ProM package baseminers 6.10.11 Does not discover models, as these models are way too big (more than 10,000 transitions) to be classified comfortably. Therefore, this algorithm uses a log-based classifyer. Classify.bat Scripts/Classify.txt Results This shows that the Trace Miner has an average F-score of 34%. The fact that it scores better when OR tasks are included can be explained by Read More …

Details The Flower Miner discover the so-called base ‘flower’ model. Prom Package baseminers 6.10.3 Discovers Petri nets Discover.bat Scripts/Discover.txt Results As every trace is classified as positive, the negative accuracy will be 0, and so will be the F-score.

Details The Directly Follows Miner is a base miner that discovers directly follows relations. Basically, every activity corresponds to a state. Every transition that correspond to an activity has the state that corresponds to that activity as target state. This way, all traces can be replayed successfully in the model. ProM package baseminers 6.10.3 Discovers Petri nets Discover.bat Scripts/Discover.txt Results This shows that the simple directly-follows miner scores 66% on the F-score, and 79% if there is no noise.

Details ProM package alphaminer-6.9.78. Discovers Petri nets. Discover.bat Scripts/Discover.txt Results This graph shows the F-score values for different configurations. This indicates that this discovery algorithm has severe issues in general, among which optional tasks. This graph shows the F-score values for the configurations without optional tasks and with optional tasks. For example, it shows that if there are no optional tasks the F-score may be 100% (4 logs), but adding optional tasks results in a drop of the F-score to Read More …

Details Fodina-2019-06-17, using ProM. Discovers Petri nets. Discover.bat Scripts/Discover.txt Results This graph shows the F-score values for different configurations. This indicates that this discovery algorithm has some issues with routing tasks. This graph shows the F-score values for the situation without routing tasks and with routing tasks. For example, it shows that if there are no routing tasks the F-score may be 100%, but adding routing tasks results in a drop of the F-score to 0% in many cases.

Details ProM package hybridilpminer-6.10.154. Discovers Petri nets. Discover.bat Scripts/Discover.txt Results This graph shows the F-score values for different configurations. This indicates that this discovery algorithm has some issues with optional tasks. This graph shows the F-score values for the configurations without optional tasks and with optional tasks. For example, it shows that if there are no optional tasks the F-score may be 100%, but adding optional tasks often results in a drop of the F-score.

Details ProM package inductiveminerdeprecated-6.10.64. Discovers Petri nets. Discover.bat Scripts/Discover.txt Results This graph shows the F-score values for different configurations. This indicates that this discovery algorithm has some issues with loops. This graph shows the F-score values for theconfigurations without loops, with simple loops, and with complex loops. For example, it shows that if there are no loops the F-score may be 100% (4 logs), but adding loops (whether simple or complex) results in a drop of the F-score.

Details ProM package inductiveminerdeprecated-6.10.64. Discovers Petri nets. Discover.bat Scripts/Discover.txt Results This graph shows the F-score values for different configurations. This indicates that this discovery algorithm has some issues with loops. This graph shows the F-score values for the configurations without loops and with loops (simple or complex). For example, it shows that if there are no loops the F-score may be 100% (4 logs), but adding loops (either simple or complex) results in a drop of the F-score.

Details ProM package logskeleton-6.10.93. Discovers log skeletons. Discover.bat Scripts/Discover.txt Results This graph shows the F-score values for different configurations. This indicates that this discovery algorithm has some issues with noise. This graph shows the F-score values for the configurations without noise and with noise. For example, it shows that if there is no noise the F-score may be 100% and is at least 80%, but adding noise results in a drop of the F-score to below 75%.

Details ProM package logskeleton-6.10.93. Discovers log skeletons allowing for 5% of noise. Discover.bat Scripts/Discover.txt Results This graph shows the F-score values for different configurations. This indicates that this discovery algorithm has some issues with loops. This graph shows the average F-score values for the configurations without loops and with loops (simple or complex). For many logs without loops it holds that introducing simple loops lowers the F-score, and introducing complex loops lowers it even further.