Automatic Identification of Bottleneck Tasks for Business Process Management using Fusion-based Text Clustering

Process bottlenecks identification and its root cause analysis using fusion-based clustering and knowledge graph

“…In topic clustering, based on previous experience, the number of design stages in this design process is usually between five and nine, so the cluster number was selected from 5 to 9. Another parameter, the fusion coefficient parameter was selected from 0.2 to 0.9 [40]. To determine optimal K and , three measures evaluation metrics: silhouette-score (S score), Calinski-Harabaz score (CH score), and Davies-Bouldin score (DBI score) [45], are utilised.…”

“…Considering that the distance function significantly affects the clustering accuracy in K-means, the BTMDW (Biterm topic model with dynamic window)-Fusion K-means clustering-based bottleneck identification algorithm was proposed [40]. Process documents are collected and preprocessed and then represented as vectors.…”

“…We briefly review hyperbolic geometry. Hyperbolic geometry has constant negative curvature and belongs to non-Euclidean geometry [40].…”