Zero-inflated poisson and negative binomial regressions for technology analysis

Jong-Min Kim, Sunghae Jun

Research output: Contribution to journalArticlepeer-review

11 Scopus citations


Technology analysis is to understand target technology by analyzing diverse information of developed technologies. Using the results of technology analysis, we can perform the technology management such as technology forecasting, technological innovation, and technology valuation for research and development (RandD) planning. In addition, the RandD planning is built upon in order to improve technological competitiveness of a company. Patent analysis is a popular approach to technology analysis. Many researches on patent analysis have been done because patent documents contain diverse and complete information on developed technology. However, the documents are not suitable for patent analysis based on statistics. So, in much of the work on patent data analysis, the researchers transformed the patent documents into structured data using text mining techniques. Generally, the structured data set has a sparsity problem, that is, most elements of the data are zero valued. The existing researches in patent analysis have not considered this zero-inflated problem, but it places serious limits on performance when we analyze the patent data. In this paper, to overcome this problem, we propose a methodology for patent analysis using zero-inflated Poisson and negative binomial regressions. We apply the proposed methodology based on zero-inflated Poisson and negative binomial regression models to Apple's technology analysis.

Original languageEnglish (US)
Pages (from-to)431-448
Number of pages18
JournalInternational Journal of Software Engineering and its Applications
Issue number12
StatePublished - Jan 1 2016


  • Apple patent
  • Patent data analysis
  • Technology analysis
  • Zero-inflated negative binomial model
  • Zero-inflated poisson model
  • Zero-inflated problem


Dive into the research topics of 'Zero-inflated poisson and negative binomial regressions for technology analysis'. Together they form a unique fingerprint.

Cite this