Incident Clustering in the Warehouse Workspaces by Using Text Mining

Abstract

Incidents can occur anytime and anywhere whether intentional or unintentional, especially in the industrial area. To prevent the incident, an analysis is needed to find out the patterns that formed based on the information from incident reports. The goals of company are to be able to identify and take action to deal quickly if some incident happens. Text Mining is done by applying the bigram technique to form the incident patterns. The analytical method used is K-Means Clustering and Hierarchical Clustering. In addition, a feature selection is also used by applying the Genetic Algorithm method to obtain optimal features. The results obtained state that the feature selection process is very influential on the formation of incident clusters. When compared, the K-Means Clustering and Hierarchical Clustering methods have different effects on each warehouse sector. The best results in the Sector of Life Style and Sector of Technology are formed by using the K-Means Clustering method while in the Sector of Consumer, Retail and SPL the best cluster results are obtained based on the Hierarchical Clustering method.

Mapping the Provincial Food Security Conditions in Indonesia Using Cluster Ensemble-Based Mixed Data Clustering-Robust Clustering with Links (CEBMDC-ROCK)

“…After getting the best number of clusters, the next step is to determine the best clustering algorithm for numerical data based on the ICD Rate value [40]. The ICD Rate formula is shown in Equation 12.…”