專利趨勢變化探勘之研究

Abstract

在企業競爭環境中，專利既是保護研發成果的防禦武器也是牽制競爭對手的攻擊武器。專利說明書蘊含豐富的技術資訊與研發趨勢，而專利分析是將「專利資訊」轉換成「競爭情報」的利器。現有的專利分析方法，只能將多個時間點的專利地圖以人工判別圖形中專利隱含趨勢，缺乏自動化的方法以找出專利不同時間點的變化趨勢。本研究提出一個探勘專利趨勢的方法，運用資料探勘之關連規則變化探勘，發掘專利資訊中隱含的趨勢變化，協助企業進行專利佈局與專利管理。本研究以半導體產業為實驗對象，使用美國專利商標局專利資料進行實驗分析，實驗結果顯示本研究所提出的方法可探勘出產業專利趨勢變化與同業競爭者專利趨勢變化，所發掘之專利趨勢變化可支援企業之專利佈局與管理。

Possessing patents is not only a defensive way to protect the achievements of research and development but also an offensive tool to impede the competitors. Patent analysis is a useful way to transform the information inside the patent documents into competitive information. Existing methods of patent analysis can only plot the quantitative information into the patent maps from different time segments, while the changes of pattern trends need to be investigated manually. This work proposes a mining approach to discover the pattern trends from the historical patent information. Association Rule Mining and Change Mining are adopted in patent analysis to discover the changes of pattern trends which can be used further to support strategy planning and management of pattern in enterprises.Experiments were conducted by applying the proposed approach to the semiconductor industry using the patent data collected from the United States Patent and Trade Office. This work presents the discovered changes of patent trends of both the industry and the key competitors in the industry. The result is useful to support strategy planning and management of pattern in enterprises.