Organizations are taking advantage of "data-mining" techniques to leverage the vast amounts of data captured as they process routine transactions. Data-mining is the process of discovering hidden structure or patterns in data. However several of the pattern discovery methods in datamining systems have the drawbacks that they discover too many obvious or irrelevant patterns and that they do not leverage to a full extent valuable prior domain knowledge that managers have. This research addresses these drawbacks by developing ways to generate interesting patterns by incorporating managers' prior knowledge in the process of searching for patterns in data. Specifically we focus on providing methods that generate unexpected patterns with respect to managerial intuition by eliciting managers' beliefs about the domain and using these beliefs to seed the search for unexpected patterns in data. Our approach should lead to the development of decision support systems that provide managers with mor...

We consider the problem of finding rules relating patterns in a time series to other patterns in that series, or patterns in one series to patterns in another series. A simple example is a rule such as "a period of low telephone call activity is usually followed by a sharp rise in call volume". Examples of rules relating two or more time series are "if the Microsoft stock price goes up and Intel falls, then IBM goes up the next day," and "if Microsoft goes up strongly for one day, then declines strongly on the next day, and on the same days Intel stays about level, then IBM stays about level." Our emphasis is in the discovery of local patterns in multivariate time series, in contrast to traditional time series analysis which largely focuses on global models. Thus, we search for rules whose conditions refer to patterns in time series. However, we do not want to define beforehand which patterns are to be used; rather, we want the patterns to be formed from the data in t...

. Learning multiple descriptions for each class in the data has been shown to reduce generalization error but the amount of error reduction varies greatly from domain to domain. This paper presents a novel empirical analysis that helps to understand this variation. Our hypothesis is that the amount of error reduction is linked to the "degree to which the descriptions for a class make errors in a correlated manner." We present a precise and novel definition for this notion and use twenty-nine data sets to show that the amount of observed error reduction is negatively correlated with the degree to which the descriptions make errors in a correlated manner. We empirically show that it is possible to learn descriptions that make less correlated errors in domains in which many ties in the search evaluation measure (e.g. information gain) are experienced during learning. The paper also presents results that help to understand when and why multiple descriptions are a help (irrelevant attribute...

We analyse the biases of eleven measures for estimating the quality of the multi-valued attributes. The values of information gain, J- measure, gini-index, and relevance tend to linearly increase with the number of values of an attribute. The values of gain-ratio, distance measure, Relief , and the weight of evidence decrease for informative attributes and increase for irrelevant attributes. The bias of the statistic tests based on the chi-square distribution is similar but these functions are not able to discriminate among the attributes of different quality. We also introduce a new function based on the MDL principle whose value slightly decreases with the increasing number of attribute's values. 1 Introduction In top down induction of decision trees various impurity functions are used to estimate the quality of attributes in order to select the "best" one to split on. However, various heuristics tend to overestimate the multi-valued attributes. One possible approach to this proble...

The field of knowledge discovery in databases, or "Data Mining", has received increasing attention during recent years as large organizations have begun to realize the potential value of the information that is stored implicitly in their databases. One specific data mining task is the mining of Association Rules, particularly from retail data. The task is to determine patterns (or rules) that characterize the shopping behavior of customers from a large database of previous consumer transactions. The rules can then be used to focus marketing efforts such as product placement and sales promotions. Because early algorithms required an unpredictably large number of IO operations, reducing IO cost has been the primary target of the algorithms presented in the literature. One of the most recent proposed algorithms, called PARTITION, uses a new TID-list data representation and a new partitioning technique. The partitioning technique reduces IO cost to a constant amount by processing one datab...

Many learning algorithms form concept descriptions composed of clauses, each of which covers some proportion of the positive training data and a small to zero proportion of the negative training data. This paper presents a method using likelihood ratios attached to clauses to classify test examples. One concept description is learned for each class. Each concept description competes to classify the test example using the likelihood ratios assigned to clauses of that concept description. By testing on several artificial and "real world" domains, we demonstrate that attaching weights and allowing concept descriptions to compete to classify examples reduces an algorithm's susceptibility to noise.

We introduce and study a novel genre of optimization problems, which we call segmentation problems. Our motivation, in part, is the development of a framework for evaluating certain data mining and clustering operations in terms of their utility in decision-making. For any classical optimization problem, the corresponding segmentation problem seeks to partition a set of cost vectors into several segments, so that the overall cost is optimized. This framework contains a number of standard combinatorial clustering problems as special cases, and many segmentation problems turn out to be MAXSNP-complete even when the corresponding "un-segmented" version is easy to solve. We develop approximation algorithms for two natural and interesting problems in this class --- the HYPERCUBE SEGMENTATION PROBLEM and the CATALOG SEGMENTATION PROBLEM --- and present a general greedy scheme, which can be specialized to approximate a large class of segmentation problems. Finally, we indicate some connection...

Abstract. Although successful in medical diagnostic problems, inductive learning systems were not widely accepted in medical practice. In this paper two di erent approaches to machine learning in medical appli-cations are compared: the system for inductive learning of decision trees Assistant, and the naive Bayesian classi er. Both methodologies were tested in four medical diagnostic problems: localization of primary tumor, prognostics of recurrence of breast cancer, diagnosis of thyroid diseases, and rheumatology. The accuracy of automatically acquired diagnostic knowledge from stored data records is compared and the interpretation of the knowledge and the explanation ability of the classi cation process of each system is discussed. Surprisingly, thenaiveBayesian classi er is superior to Assistant in classi cation accuracy and explanation ability, while the interpretation of the acquired knowledge seems to be equally valuable. In ad-dition, two extensions to naive Bayesian classi er are brie y described: dealing with continuous attributes, and discovering the dependencies among attributes.

Knowledge discovery in databases, also known as data mining, is the efficient discovery of previously unknown, valid, novel, potentially useful, and understandable patterns in large databases. It encompasses many different techniques and algorithms which differ in the kinds of data that can be analyzed and the form of knowledge representation used to convey the discovered knowledge. An important problem in the area of data mining is the development of effective measures of interestingness for ranking the discovered knowledge. In this report, we provide a general overview of the more successful and widely known data mining techniques and algorithms, and survey seventeen interestingness measures from the literature that have been successfully employed in data mining applications. 1 1

Mining for association rules is considered an important data mining problem. Many different variations of this problem have been described in the literature. In this paper we introduce the problem of mining for negative associations. A naive approach to finding negative associations leads to a very large number of rules with low interest measures. We address this problem by combining previously discovered positive associations with domain knowledge to constrain the search space such that fewer but more interesting negative rules are mined. We describe an algorithm that efficiently finds all such negative associations and present the experimental results. 1 Introduction Wide spread use of computers in business operations and the availability of cheap storage devices have led to an explosive growth in the amount of data gathered and stored by most business organizations today. There has been a trend in recent years to search for interesting patterns in the data and use them for improve...