Search

Data mining focuses on the development of methods and algorithms for such various tasks as classification, clustering, rule induction, and discovery of associations. In the database fields, the view of data mining as advanced querying has recently stimulated much research into the development of data mining query languages. In the fields of machine learning, inductive logic programming has broadenes its scope towards extending standard data mining tasks from the usual attribute-value setting to a multi-relational setting. After a concise description of data mining, the contribution of logic to both fields is discussed. At the end, we indicate the potential use of logic for unifying different existsing data mining formalisms.

The key point of this paper is that, in frequent pattern mining, the most appropriate way of exploiting monotone constraints in conjunction with frequency is to use them in order to reduce the problem input together with the search space. Following this intuition, we introduce ExAMiner, a level-wise algorithm which exploits the real synergy of anti-monotone and monotone constraints: the total benefit is greater than the sum of the two individual benefits. ExAMiner generalizes the basic idea of the preprocessing algorithm ExAnte, embedding such ideas at all levels of an Apriori-like computation. The resulting algorithm is the generalization of the Apriori algorithm when a conjunction of monotone constraints is conjoined to the frequency anti-monotone constraint. Experimental results confirm that this is, so far, the most efficient way of attacking the computational problem in analysis.

An abstract is not available

Constraint pushing techniques have been proven to be effective in reducing the search space in the frequent pattern mining task, and thus in improving efficiency. But while pushing anti-monotone constraints in a level-wise computation of frequent itemsets has been recognized to be always profitable, the case is different for monotonic constraints. In fact, monotone constraints have been considered harder to push in the computation and less effective in pruning the search space. In this paper, we show that this prejudice is ill-founded and introduce ExAnte, a preprocessing data reduction algorithm which reduces dramatically both the search space and the input dataset in constrained frequent pattern mining. Experimental results show a reduction of orders of magnitude, thus enabling a much easier mining task. ExAnte can be used as a mining preprocessor with any constraint pattern mining algorithm.

Finding information using Web search engines is not always successful. When search results are presented in a ranked list, users are often compelled to sift through a long list of snippets to find the information they are looking for. This paper presents a versatile system to reorganize search results into a partition of homogeneous document clusters using Data Mining techniques. The purpose is to help users to easily browse through the set of retrieved documents, by focusing on clusters whose characterizing keywords are directly pertinent to the query. We experienced interesting results by applying our techniques on snippets only, i.e., by running our application on the client side, 'outside' of the search engine. For general queries, the obtained clusters are usually natural collections of homogeneous documents, and often documents in the same cluster occur in distant positions in the ranked list returned by the search engine.

Current applications of data mining techniques highlight the need for flexible knowledge discovery systems, capable of supporting the user in specifying and re.ning mining objectives, combining multiple strategies, and de.ning the quality of the extracted knowledge. A key issue is the de.nition of Knowledge Discovery Support Environment, i.e., a query system capable of obtaining, maintaining, representing and using high level knowledge in a uni.ed framework. This comprises representation and manipulation of domain knowledge, extraction and manipulation of new knowledge and their combination.

An abstract is not available

An abstract is not available

Nondeterministic extensions are needed in logic-based languages, such as first-order relational languages and Datalog, to enhance their expressive power and support the efficient formulation of low-complexity problems and database queries. In this paper, we study the semantics and expressive power of the various nondeterministic constructs proposed in the past, including various versions of the choice operator and the witness operator. The paper develops a model-theoretic semantics, a fixpoint semantics, and an opera- tional semantics for these constructs, and characterizes their power of expressing deterministic and nondeterministic queries. The paper presents various soundness and completeness results and establishes an expressiveness hierarchy that correlates the various operators with each other and with other constructs such as negation and fixpoint.

An abstract is not available.

An abstract is not available.

An abstract is not available.

This paper presents a logic database language with elementary data mining mechanisms to model the relevant aspects of knowledge discovery , and to provide support for both the interative and interactive features of the knowledge discovery proccess

An abstract is not available.

We consider in this paper an extension of Datalog withmech- anisms for non-monotonic and non-deterministic reasoning and a simple form of temporal reasoning, which we refer to as Datalog++. First, we show how with this logic database language is possible to express prob- lems in heterogeneous domains, such as operation research and concur- rent programming. Second, we provide a methodology for the verifica- tion of Datalog++ programs, based on the declarative semantics, which is able to handle both atemporal and temporal properties.

An abstract is not available.

We report here the vision of the Pisa KDD Laboratory, a joint effort of University of Pisa and Italian National Research Council - a research group operating in the area of data mining and knowledge discovery in databases. We first advocate the need of a logic-based knowledge discovery support environment, capable of integrating knowledge extraction and knowledge manipulation. It is our conviction that a high degree of flexibility and expressiveness is needed to support the process of knowledge discovery in databases in most challenging data analysis applications. The KDD Laboratory is developing a data mining query language, which deals effectively and uniformly with data preparation, model extraction and model evaluation and analysis, thus providing a powerful formalism where methodologies for classes of challenging applications can be conveniently designed.

We present the notion of Iterative user-defined aggregates as an extension of the notion of user-defined aggregates in deductive databases. Such an extension provides a versative mechanism for defining complex aggregation functions, that are not definable as distributive aggregates. As a result, we show how such a mechanism can be applied to the specification of complex data mining tasks as user-defined aggregates. The resulting formalism provides a flexible way to customize, tune and reason on both the evaluation functions and the extracted knowledge.

Questo rapporto tecnico descrive una serie di esperimenti sistematici di analisi dei dati del Data Warehouse Piano telematico Calabria mediante tecniche di mining

An abstract is not available.

A Logic-based query language with built-in data mining primitives is presented. The aim is to have a language where knowledge extra ction and knowledge manipulation are integrated. Its flexibility and expressiveness in supporting the process of knowledge discovery in databases is illustrated presenting a market-basket analysis application. By focussing on association rules, we show that the query language deals effectively and uniformly with data preparation, rule extraction, analysis and construction of business rules. The implementation of the language is based on an architecture that allows a convenient compromise between tight and loose coupling of database querying and data mining.

Datasift, a prototype system for the analysis of supermarket sales data based on data mining techniques, is presented. In market basket analysis, mining knowledge on customer behavior, which is actually useful to support marketing actions, is a difficult task, which requires non-trivial methods of employing and combining the data mining tools. To this purpose, we propose an architecture that integrates the deductive capabilities of a logic-based database language, LDL++, with the inductive capabilities of diverse data mining algorithms and tools, notably association rules. This paper presents an application of the integrated system to market basket analysis, and illustrates the high degree of expressiveness reached in dealing with high-level business rules.

An abstract is not available.

An abstract is not available.

An abstract is not available

We consider in this paper an extension of Datalog with mechanisms for temporal, nonmonotonic and nondeterministic reasoning, which we refer to as Datalog++. We study its semantics, and show how iterated fixpoint and stable model semantics can be combined to the purpose of clarifying the interpretation of Datalog++ programs, and supporting their efficient execution. On this basis, the design of appropriate optimization techniques for Datalog++ is also discussed.