Typical short-term remedy knowledge mining for product quality problem-solving based on bipartite graph clustering

Zhaoguang Xu, Yanzhong Dang, Zhongzhao Zhang, Jingfeng Chen

Research output: Contribution to journalArticlepeer-review

8 Scopus citations


The increasing demand for high product quality by consumers poses new challenges to the efficiency and effectiveness of manufacturers’ quality problem-solving. Solving problems based on personal experience makes problem-solving inefficient and ineffective. However, the data recorded during the problem-solving process can offer valuable experiential knowledge for problem-solvers. In this study, we propose a bipartite graph clustering method for discovering the knowledge of short-term remedies, which is a type of solution, from past quality problem-solving data. In this method, several clustering algorithms are compared, and the K-means algorithm is selected to cluster quality problems into typical problem clusters. A novel two-stage clustering method based on verb and noun clustering is then developed to construct typical short-term remedy clusters. Based on the clustering result, the relationship between problem clusters and short-term remedy clusters is generated. A reasoning method for extracting short-term remedy knowledge to solve new problems is introduced, and quality problem-solving data on an automobile manufacturer are used to carry out a case study. Tools such as Gephi and a prototype system are applied to provide “problem cluster–short-term remedy cluster” knowledge. Problem-solvers can use this knowledge to quickly address new problems, thereby improving the efficiency and effectiveness of product quality problem-solving.

Original languageEnglish
Article number103277
JournalComputers in Industry
StatePublished - Nov 2020
Externally publishedYes


  • Bipartite graph
  • Clustering
  • Know-how
  • Quality problem-solving
  • Text mining


Dive into the research topics of 'Typical short-term remedy knowledge mining for product quality problem-solving based on bipartite graph clustering'. Together they form a unique fingerprint.

Cite this