新聞事件之變化探勘以支援決策制定

Abstract

企業目前面臨的是一個變動快速的環境，組織內外部皆都受到許多不確定因素影響，因此追蹤、監督與管理企業有關的新舊事件，進而掌握環境演變對企業而言相當重要。然而現有研究僅著重於新聞事件中新事件的追踨與偵測，無法有效辨認事件變化趨勢。本研究提出一個發掘事件變化的方法，結合資料探勘方法中關連式規則與概念階層，從大量且具有時間性的新聞資料中，針對特定主題和事件進行分析與研究，藉由五種關連規則的變化類型:Emerging Patterns、Unexpected Condition、Unexpected Consequence、Added Rule和Perished Rule來找尋事件資料的演變，以發掘事件的變動趨勢與產業脈動，將事件變化資訊提供予決策者參考，協助企業達成決策制定的目的。

Business environment, in which an enterprise operates locally or globally, has been changing at an unprecedented rate; and business decisions are affected by many uncertain factors from both the inside and outside of the enterprise. Therefore, it is important for an enterprise to track, supervise and manage the existing and forthcoming events pertinent to its business. Recently, many researches have been done on event tracking and detection; however, identifying event changes has not been considered. In this research, we provide a method of combining association rule mining and concept hierarchy for discovering event changes from news events. We focus on specific topics and events to find out trends and changes in news data according to five types of association rules changes, such as emerging pattern, unexpected condition, unexpected consequence, added rule and perished rule. The information of event changes can then be provided to the decision makers for decision support.