Tomasz Łukaszewski

The role of semantics in mining frequent patterns from knowledge bases in description logics with rules

We propose a new method for mining frequent patterns in a language that combines both Semantic Web ontologies and rules. In particular, we consider the setting of using a language that combines description logics (DLs) with DL-safe rules. This setting is important for the practical application of data mining to the Semantic Web. We focus on the relation of the semantics of the representation formalism to the task of frequent pattern discovery, and for the core of our method, we propose an algorithm that exploits the semantics of the combined knowledge base. We have developed a proof-of-concept data mining implementation of this. Using this we have empirically shown that using the combined knowledge base to perform semantic tests can make data mining faster by pruning useless candidate patterns before their evaluation. We have also shown that the quality of the set of patterns produced may be improved: the patterns are more compact, and there are fewer patterns. We conclude that exploiting the semantics of a chosen representation formalism is key to the design and application of (onto-)relational frequent pattern discovery methods.

Faster Frequent Pattern Mining from the Semantic Web

In this paper we propose a method for frequent pattern discovery from the knowledge bases represented in OWL DLP. OWL DLP, known also as Description Logic Programs, is the intersection of the expressivity of OWL DL and Logic Programming. Our method is based on a special form of a trie data structure. A similar structure was used for frequent pattern discovery in classical and relational data mining settings giving significant gain in efficiency. Our approach is illustrated on the example ontology.

On Reducing Redundancy in Mining Relational Association Rules from the Semantic Web

In this paper we discuss how to reduce redundancy in the process and in the results of mining the Semantic Web data. In particular, we argue that the availability of the domain knowledge should not be disregarded during data mining process. As the case study we show how to integrate the semantic redundancy reduction techniques into our approach to mining association rules from the hybrid knowledge bases represented in OWL with rules.

Frequent pattern discovery from OWL DLP knowledge bases

The Semantic Web technology should enable publishing of numerous resources of scientific and other, highly formalized data on the Web. The application of mining these huge, networked Web repositories seems interesting and challenging. In this paper we present and discuss an inductive reasoning procedure for mining frequent patterns from the knowledge bases represented in OWL DLP. OWL DLP, also known as Description Logic Programs, lies at the intersection of the expressivity of OWL DL and Logic Programming. Our method is based on a special trie data structure inspired by similar, efficient structures used in classical and relational data mining settings. Conjunctive queries to OWL DLP knowledge bases are the language of frequent patterns.

Towards discovery of frequent patterns in description logics with rules

This paper follows the research direction that has received a growing interest recently, namely application of knowledge discovery methods to complex data representations. Among others, there have been methods proposed for learning in expressive, hybrid languages, combining relational component with terminological (description logics) component. In this paper we present a novel approach to frequent pattern discovery over the knowledge base represented in such a language, the combination of the basic subset of description logics with DL-safe rules, that can be seen as a subset of Semantic Web Rule Language. Frequent patterns in our approach are represented as conjunctive DL-safe queries over the hybrid knowledge base. We present also an illustrative example of our method based on the financial dataset.

Controlling the prediction accuracy by adjusting the abstraction levels

The predictive accuracy of classifiers is determined among others by the quality of data. This important property of data is strongly affected by such factors as the number of erroneous or missing attributes present in the dataset. In this paper we show how those factors can be handled by introducing the levels of abstraction in data definition. Our approach is especially valuable in cases where giving the precise value of an attribute is impossible for a number of reasons as for example lack of time or knowledge. Furthermore, we show that increasing the level of precision for an attribute significantly increase predictive accuracy, especially when it is done for the attribute with high information gain.

Classification with test costs and background knowledge

Abstract We propose a novel approach to the problem of the classification with test costs understood as costs of obtaining attribute values of classified examples. Many existing approaches construct classifiers in order to control the tradeoff between test costs and the prediction accuracy (or misclassification costs). The aim of the proposed method is to reduce test costs while maintaining of the prediction accuracy of a classifier. We assume that attribute values are represented at different levels of abstraction and model domain background knowledge. Our approach sequentially explores these levels during classification – in each iteration it selects and conducts a test that precises the representation of a classified example (i.e., acquires an attribute value), invokes a naive Bayes classifier for this new representation and checks the classifier’s outcome to decide whether this iterative process can be stopped. The selection of the test in each iteration takes into account the possible improvement of the prediction accuracy and the cost of this test. We show that the prediction accuracy obtained for classified examples represented precisely (i.e., when all the tests have been conducted and all specific attribute values have been acquired) can be achieved for a much smaller number of tests (i.e., when not all specific attribute values have been acquired). Moreover, we show that without levels of abstraction and with uniform test costs our method can be used for selecting features and it is competitive to popular feature selection schemes: filter and wrapper.

Sequential Classification by Exploring Levels of Abstraction

Abstract In the paper we describe a sequential classification scheme that iteratively explores levels of abstraction in the description of examples. These levels of abstraction represent attribute values of increasing precision. Specifically, we assume attribute values constitute an ontology (i.e., attribute value ontology) reflecting a domain-specific background knowledge, where more general values subsumes more precise ones. While there are approaches that consider levels of abstraction during learning, the novelty of our proposal consists in exploring levels of abstraction when classifying new examples. The described scheme is essential when tests that increase precision of example description are associated with costs – such a situation is often encountered in medical diagnosis. Experimental evaluation of the proposed classification scheme combined with ontological Bayes classifier (i.e., a naive Bayes classifier expanded to handle attribute value ontologies) demonstrates that the classification accuracy obtained at higher levels of abstraction (i.e., more general description of classified examples) converges very quickly to the classification accuracy for classified examples represented precisely. This finding indicates we should be able to reduce the number of tests and thus limit their cost without deterioration of the prediction accuracy.

A Study of the SEMINTEC Approach to Frequent Pattern Mining

This paper contains the experimental investigation of an approach, named SEMINTEC, to frequent pattern mining in combined knowledge bases represented in description logic with rules (so-called \({\mathcal DL}\)-safe ones). Frequent patterns in this approach are the conjunctive queries to a combined knowledge base. In this paper, first, we prove that the approach introduced in our previous work for the DLP fragment of description logic family of languages, is also valid for more expressive languages. Next, we present the experimental results under different settings of the approach, and on knowledge bases of different sizes and complexities.

An Evolutionary Algorithm for Bayesian Network Triangulation

The problem of triangulation (decomposition) of Bayesian networks is considered. Triangularity of a Bayesian network is required in a general evidence propagation scheme on this network. Finding an optimal triangulation is NP-hard. A local search heuristic based on the idea of evolutionary algorithms is presented. The results obtained using existing and proposed approaches are compared on a basis of a computational experiment.The problem of triangulation (decomposition) of Bayesian networks is considered. Triangularity of a Bayesian network is required in a general evidence propagation scheme on this network. Finding an optimal triangulation is NP-hard. A local search heuristic based on the idea of evolutionary algorithms is presented. The results obtained using existing and proposed approaches are compared on a basis of a computational experiment.