Business Process Models Clustering Based on Multimodal Search, K-means, and Cumulative and No-Continuous N-Grams

Abstract

Due to the large volume of process repositories, finding a particular process may become a difficult task. This paper presents a method for indexing, search, and grouping business processes models. The method considers linguistic and behavior information for modeling the business process. Behavior information is described using cumulative and nocontinuous n-grams. Grouping method is based on k-means algorithm and suffix arrays to define labels for each group. The clustering approach incorporates mechanisms for avoiding overlapping and improve the homogeneity of the created groups using the K-means algorithm. Obtained results outperform the precision, recall and F-measure of previous approaches. Index Terms-Clustering,business process models, multimodal search, cumulative and no-continuous n-grams.

A business process clustering algorithm using incremental covering arrays to explore search space and balanced Bayesian information criterion to evaluate quality of solutions

“…This set was collaboratively created by 59 experts. Furthermore, the results of the ICAClusterBP algorithm were compared against the results of four state-of-the-art algorithms; CAClusterBP presented in [16] which used an ICA (5, 40; 2, 20, 5) and Euclidean measure to define the similarity of the BPs within each group; the HC algorithm [32], which uses cosine coefficient to measure the similarity between two process models and implements a hierarchical agglomerative algorithm for clustering; k-meansBP that uses the k-means clustering algorithm (adapted for use in BPs) and a multimodal search model based on cosine distance [41]; and finally, GroupBPFuzzy presented in [8]—this algorithm uses a multimodal method to retrieve BPs from a repository and an incremental algorithm to group retrieved BPs. Groups are created using a fuzzy-based function that measures similarity between two BPs according to the number of common elements between them.…”

“…Elsewhere, Ordoñez et al use k-means to group BPs retrieved from a repository through a multimodal search system that integrates textual and structural information [41]. Structural information is represented as codebooks in the form of text strings that simulate the formation of non-continuous and cumulative N-grams.…”