應用資料探勘分類技術於專利分析之研究

A Study of Applying Data Mining Classification Techniques to Patent Analysis

李駿翔(Chun-Hsiang Lee)

指導教授 ： 劉士豪

中原大學/商學院/資訊管理研究所/碩士(2003年)

https://doi.org/10.6840/cycu200300636

摘要

隨著政府對於智慧財產權的重視，以及廠商對於研發成果保護的殷切需要，再加上專利可說是智慧結晶最為具體之表現，因此專利分析遂成為各家廠商的主要課題。專利分析是將專利「資訊」加值轉換成有用的專利「情報」的重要過程，一方面可藉由分析找出自身的競爭優勢，研發出自身獨特的產品；另一方面，可以瞭解競爭者的專利範圍，以免誤觸他人專利地雷。根據世界智慧財產權組織的報告指出，專利說明書中含有90%~95%之研發成果，善加利用專利資訊，不但可縮短60%的研發時間，更可節省將近40%的研究經費，證明專利文件對於科技研究人員而言，是最具參考價值之重要技術資訊。
然而對於企業來說，專利分析的成本相當高昂，其原因在於專利分析所需花費的時間太過冗長，且分析人員需具有相當程度之專業知識，以至於所需人力成本攀高，此外，國內在專利分析系統工具上也非常的缺乏，使得專利分析仍需完全仰賴人力。因此，本研究利用資料探勘之技術於專利分析當中，輔助專利分析當中，最為基本之專利文件的自動分類當中，並藉由「基因轉殖生物」相關之專利資料進行分類測試，針對各種不同的實驗變數，找尋最佳之應用模式，且透過本研究之驗證，確立使用資料探勘技術於專利分析之可行性。
研究結果顯示，利用資料探勘於專利文件之技術分類時，有顯著之分類效果，透過研究過程當中調整階段之實驗變數配合，將可有效過濾出離群值，以增加分類之精確率，透過研究結果也發現，實際應用時將可根據不同之需求來調整實驗當中各變數之內容，以符合分析之需求。

關鍵字

專利分析 ； 資料探勘 ； 分類

並列摘要

With the government attaching importance to intellectual property right, the need of enterprise protecting the research result, and the patent begin also the most specific expression of human wisdom, patent analysis become the main subject for executive of enterprise progressively. Patent analysis is the value-added process transferring the information of patents to practical intelligence. Depending on patent analysis, enterprise can find out the self-advantage in the competition with other opponents and understand the patent claim of competitors to avoid step on the patent-landmine. According to the report of World Intellectual Property Organization (WIPO): There is 90%-95% of whole world research result can be found in the paten data, if enterprise can use the patent data in research and develop well, then it can shorten the 60% R&D time and save 40% cost. It proves the patent data is most worthy of study for the researcher in R&D department.
However, the cost of patent analysis for enterprise is fairly expensive. The reason is it needs a lot of time to analysis and analyst must have a high level of domain knowledge. Besides, there are fairly few information systems about patent analysis in Taiwan, so the patent analysis depends on human resource completely.
Therefore, this study uses the data mining techniques to assist in the first step in patent analysis - patent classification automatically. By using the patent data about “Genetically Modified Organisms, GMO” to experiment with various parameters, and find out the best practical type to verify the feasibility of applying data mining techniques to patent analysis.
Based on the main findings of the study, some discovers are listed below:
First, applying data mining to the technical classification of patents is effective. Second, by adjusting parameters in the tuning phase applying data mining can filter the outlier in patent classification. Third, it can change the configuration of various parameters to fit with the need of analysis.

並列關鍵字

patent analysis ； classification ； data mining

參考文獻

1. 許哲銘，「時間序列型態之知識探索」，國立中山大學資訊管理研究所碩士論文，1999年。

1. Andersen, B., “The evolution of technological trajectories 1890-1990,” Structural Change and Economic Dynamics, 1998, Vol.9, pp.5-34.

2. Berrym M., J., A., Linoff, G., “Data Mining Techniques: For Marketing Sale and Customer Support,” John Wiley & Sons, Inc., 1997.

4. Cantwell, J., Janne, O., “Technological globalization and innovative centers: the role of corporate technological leadership and locational hierarchy,” Research Policy, 1999, Vol.28, pp.119-144.

7. Debackere, K, Luwel, M., and Veugelers, R., “Can technology lead to a competitive? A case study of Flanders using European patent,” Scentometrics, 1999, Vol.44, No.3, pp.379-400.