Attribute‐oriented induction (AOI) is a data analysis technique based on induction. The traditional AOI algorithm requires a threshold given by users to determine the number of output tuples. However, it is not easy to set an appropriate tuple threshold, and there is usually noise contained in a dataset. The traditional AOI algorithm can only generate a summary output of a fixed size, but it cannot guarantee that all generalized tuples have sufficient specificity and representativeness. In this article, a new AOI method is proposed to make up for the shortcomings. We introduce the concept of cost to measure the loss of accuracy due to attribute ascension. We also propose two algorithms based on the hierarchical clustering method. By setting cost constraints on each generalized tuple, our method can generate accurate output while eliminating noise, and help users get more informative and clearer results.

中文翻译：

---

通过面向属性的归纳方法挖掘特定的代表性信息

面向属性的归纳（AOI）是一种基于归纳的数据分析技术。传统的AOI算法需要用户指定一个阈值来确定输出元组的数量。但是，设置适当的元组阈值并不容易，并且通常在数据集中包含噪声。传统的AOI算法只能生成固定大小的摘要输出，但不能保证所有广义元组都具有足够的特异性和代表性。在本文中，提出了一种新的AOI方法来弥补该缺点。我们引入成本的概念来度量由于属性提升而导致的准确性损失。我们还提出了两种基于层次聚类方法的算法。通过在每个广义元组上设置成本约束，我们的方法可以生成准确的输出，同时消除噪声，