In the feature subset selection problem, a learning algorithm is faced with the problem of selecting a relevant subset of features upon which to focus its attention, while ignoring the rest. To achieve the best possible performance with a particular learning algorithm on a particular training set, a feature subset selection method should consider how the algorithm and the training set interact. We explore the relation between optimal feature subset selection and relevance. Our wrapper method searches for an optimal feature subset tailored to a particular algorithm and a domain. We study the strengths and weaknesses of the wrapper approach and show a series of improved designs. We compare the wrapper approach to induction without feature subset selection and to Relief, a filter approach to feature subset selection. Significant improvement in accuracy is achieved for some datasets for the two families of induction algorithms used: decision trees and

Instance-based learning techniques typically handle continuous and linear input values well, but often do not handle nominal input attributes appropriately. The Value Difference Metric (VDM) was designed to find reasonable distance values between nominal attribute values, but it largely ignores continuous attributes, requiring discretization to map continuous values into nominal values. This paper proposes three new heterogeneous distance functions, called the Heterogeneous Value Difference Metric (HVDM), the Interpolated Value Difference Metric (IVDM), and the Windowed Value Difference Metric (WVDM). These new distance functions are designed to handle applications with nominal attributes, continuous attributes, or both. In experiments on 48 applications the new distance metrics achieve higher classification accuracy on average than three previous distance functions on those datasets that have both nominal and continuous attributes. 1. Introduction Instance-Based Learning (IBL) (Aha, ...

Data mining algorithmsincluding machine learning, statistical analysis, and pattern recognition techniques can greatly improve our understanding of data warehouses that are now becoming more widespread. In this paper, we focus on classification algorithms and review the need for multiple classification algorithms. We describe a system called MLC++ , which was designed to help choose the appropriate classification algorithm for a given dataset by making it easy to compare the utility of different algorithms on a specific dataset of interest. MLC ++ not only provides a workbench for such comparisons, but also provides a library of C ++ classes to aid in the development of new algorithms, especially hybrid algorithms and multi-strategy algorithms. Such algorithms are generally hard to code from scratch. We discuss design issues, interfaces to other programs, and visualization of the resulting classifiers. 1 Introduction Data warehouses containing massive amounts of data have been b...

In this doctoral dissertation, we study three basic problems in machine learning and two new hypothesis spaces with corresponding learning algorithms. The problems we investigate are: accuracy estimation, feature subset selection, and parameter tuning. The latter two problems are related and are studied under the wrapper approach. The hypothesis spaces we investigate are: decision tables with a default majority rule (DTMs) and oblivious read-once decision graphs (OODGs).

. Instance-based learning algorithms are often faced with the problem of deciding which instances to store for use during generalization. Storing too many instances can result in large memory requirements and slow execution speed, and can cause an oversensitivity to noise. This paper has two main purposes. First, it provides a survey of existing algorithms used to reduce storage requirements in instance-based learning algorithms and other exemplar-based algorithms. Second, it proposes six additional reduction algorithms called DROP1--DROP5 and DEL (three of which were first described in Wilson & Martinez, 1997c, as RT1--RT3) that can be used to remove instances from the concept description. These algorithms and 10 algorithms from the survey are compared on 31 classification tasks. Of those algorithms that provide substantial storage reduction, the DROP algorithms have the highest average generalization accuracy in these experiments, especially in the presence of uniform class noise. ...

The nearest neighbor algorithm and its derivatives are often quite successful at learning a concept from a training set and providing good generalization on subsequent input vectors. However, these techniques often retain the entire training set in memory, resulting in large memory requirements and slow execution speed, as well as a sensitivity to noise. This paper provides a discussion of issues related to reducing the number of instances retained in memory while maintaining (and sometimes improving) generalization accuracy, and mentions algorithms other researchers have used to address this problem. It presents three intuitive noise-tolerant algorithms that can be used to prune instances from the training set. In experiments on 29 applications, the algorithm that achieves the highest reduction in storage also results in the highest generalization accuracy of the three methods.

The basic nearest neighbour classifier suffers from the indiscriminate storage of all presented training instances. With a large database of instances classification response time can be slow. When noisy instances are present classification accuracy can suffer. Drawing on the large body of relevant work carried out in the past 30 years, we review the principle approaches to solving these problems. By deleting instances, both problems can be alleviated, but the criterion used is typically assumed to be all encompassing and effective over many domains. We argue against this position and introduce an algorithm that rivals the most successful existing algorithm. When evaluated on 30 different problems, neither algorithm consistently outperforms the other: consistency is very hard. To achieve the best results, we need to develop mechanisms that provide insights into the structure of class definitions. We discuss the possibility of these mechanisms and propose some initial measures that could be useful for the data miner.

Despite the success of many pattern recognition systems in constrained domains, problems that involve noisy input and many classes remain difficult. A promising direction is to use several classifiers simultaneously, such that they can complement each other in correctness. This thesis is concerned with decision combination in a multiple classifier system that is critical to its success. A multiple classifier system consists of a set of classifiers and a decision combination function. It is a preferred solution to a complex recognition problem because it allows simultaneous use of feature descriptors of many types, corresponding measures of similarity, and many classification procedures. It also allows dynamic selection, so that classifiers adapted to inputs of a particular type may be applied only when those inputs are encountered. Decisions by the classifiers are represented as rankings of the class set that are derivable from the results of feature matching. Rank scores contain more ...

The basic nearest-neighbor rule generalizes well in many domains but has several shortcomings, including inappropriate distance functions, large storage requirements, slow execution time, sensitivity to noise, and an inability to adjust its decision boundaries after storing the training data. This paper proposes methods for overcoming each of these weaknesses and combines these methods into a comprehensive learning system called the Integrated Decremental Instance-Based Learning Algorithm (IDIBL) that seeks to reduce storage, improve execution speed, and increase generalization accuracy, when compared to the basic nearest neighbor algorithm and other learning models. IDIBL tunes its own parameters using a new measure of fitness that combines confidence and cross-validation (CVC) accuracy in order to avoid discretization problems with more traditional leave-one-out cross-validation (LCV). In our experiments IDIBL achieves higher generalization accuracy than other less comprehensive instance-based learning algorithms, while requiring less than onefourth the storage of the nearest neighbor algorithm and improving execution speed by a corresponding factor. In experiments on 21 datasets, IDIBL also achieves higher generalization accuracy than those reported for 16 major machine learning and neural network models.

This paper presents an inductive learning system called the Genetic Instance-Based Learning (GIBL) system. This system combines instance-based learning approaches with evolutionary computation in order to achieve high accuracy in the presence of irrelevant or redundant attributes. Evolutionary computation is used to find a set of attribute weights that yields a high estimate of classification accuracy. Results of experiments on 16 data sets are shown, and are compared with a non-weighted version of the instance-based learning system. The results indicate that the generalization accuracy of GIBL is somewhat higher than that of the non-weighted system on regular data, and is significantly higher on data with irrelevant or redundant attributes.