본문 바로가기
  • Home

A Study on Industry Information Analysis Methodology Based on Text Mining: PEST and Polarity Analysis Using Sentence Classification

  • Journal of Knowledge Information Technology and Systems
  • Abbr : JKITS
  • 2017, 12(1), pp.109-122
  • DOI : 10.34163/jkits.2017.12.1.011
  • Publisher : Korea Knowledge Information Technology Society
  • Research Area : Interdisciplinary Studies > Interdisciplinary Research
  • Published : February 28, 2017

KIM, Yoonsung 1 Ho-Chang Lee 1 Lee, Seok Kee 2 Lee, Do-Gil 1 Hangook Kim 3 You-Eil Kim 3

1고려대학교
2한성대학교
3한국과학기술정보연구원

Accredited

ABSTRACT

Today’s companies are in an environment where they have to survive in ever-increasing competition in the industry, by constantly identifying changes and trends in their industries and by periodically reflecting them in their policies and product development. For this purpose, one of the tasks that should be carried out is the analysis of industrial information. Most companies acquire industry analytical information at the cost of a large amount of time, manpower or, with the help of external professional analysts. However, since this conventional method is a somewhat heuristic and qualitative approach. The quality of these analysis results are different each time. A huge amount of industry related information is produced online in real time and when the information is reflected in the analysis as much as possible, it is required to introduce a new analytical method. In this paper, we propose a text mining methodology that extracts information from large amount of source data and automatically classifies it into each category of industry analysis framework. By constructing a sentence classifier using feature selection technique based on machine learning method, information that can be classified by indicators of universally used industry analysis framework is collected in sentence form. We performed PEST and polarity analysis by using our system and evaluated the classification accuracy of the proposed system through experiments.

Citation status

* References for papers published after 2023 are currently being built.

This paper was written with support from the National Research Foundation of Korea.