| 摘要: | 近年來由於資訊科技的發達，已有許多技術及方法被成功的發展出來，用來挖掘有用及有趣的資訊模式，包括觀念描述、關聯規則、分類與預測、叢集和演化分析等，本論文提出一種新的資訊模式，稱為典型資料模式，提供決策者對給定的資料集有更好的了解。假定給定一個包含n個物件的資料集，每個物件可以以一組屬性值來描述，典型資料模式挖掘將由資料集中，選擇出一個緊實而適合的k物件子集合，用來代表整個資料集，根據這樣的定義，本研究提出典型資料模式挖掘方法，並且以幾個真實資料集來實作，找出有用的典型資料模式。另外，由於自動化的典型資料挖掘方法無法藉助使用者的專業知識與經驗，本研究也提出動態的使用者互動式典型資料模式挖掘方法，讓使用者可以根據經驗和專業的知識操控參數，以獲得更好的典型資料模式挖掘結果，根據所提出的互動模式，本論文開發使用者互動典型資料模式挖掘系統，以挖掘資訊系統相關典型期刊，提供一個比靜態的典型資料模式挖掘更有效的方法。 Many approaches have been proposed to discover useful information patterns from databases, such as concept description, associations, sequential patterns, classification, clustering, and deviation detection. This research proposes a new type of information pattern, called typical patterns, which can provide decision makers with a better understanding of a given dataset. Suppose we are given a dataset containing n objects, each of which is described by a set of attribute values. Mining typical patterns is to select a small subset of objects, say k objects, from these n objects so that these k chosen objects are a compact and suitable representation of the original dataset. Accordingly, the Typical Patterns Mining (TPM) algorithms have been developed to mine typical patterns from databases. Also, extensive experiments have been carried out using real datasets to demonstrate the usefulness of typical patterns in practical situations. Then, although TPM is a good method to automatically determine typical patterns, it lacks ability to accommodate user’s experience and domain knowledge, which are very crucial for making decision in a dynamic business environment. Therefore, this research also develops a dynamic and interactive approach for typical pattern mining, called interactive Typical Pattern Mining (iTPM). In this approach, we accommodate users’ experiences and knowledge by allowing users to iteratively adjust the parameters during the interactive process. Then, an iTPM system is developed to mine typical journals of IS field. The results of experiments indicate that iTPM is more effective than the previous static approach. |
