Samir Elloumi

Galois connection, formal concepts and Galois lattice in real relations: application in a real classifier

In this paper, we introduce the notion of a real set as an extension of a crisp and a fuzzy set by using sequences of intervals as membership degrees, instead of a single value in [0,1]. We also propose, to extend the notion of Galois connection in a real binary relation as well as the notions of rectangular relation, formal concept and Galois lattice. We present finally a real classifier based on this mathematical foundation.

An incremental learning system for imprecise and uncertain knowledge discovery

Discovering knowledge from databases in order to classify new patterns is an interesting field for machine learning methods. Particularly, rule induction approaches constitute prominent machine learning methods that lead to avoid the disadvantages of the decision tree. The fuzzy incremental production rule (FIPR) based system is a rule induction system that generates imprecise and uncertain IF-THEN rules from data records. It allows the incremental maintenance of the knowledge base with a minimal overhead. The precision analysis with real world data sets, and the complexity analysis are used to compare this system with existing ones and to prove the usefulness of fuzzy knowledge representation.

Extension of fuzzy Galois connection for information retrieval using a fuzzy quantifier

Summary form only given. We have proposed another approach for the logical model of IR, for the evaluation of the IR implication d /spl iexcl/! RQ in a fuzzy context. We transform the initial implication into a fuzzy implication and the matching process is modelled by the extension of the fuzzy Galois connection while considering a fuzzy quantifier. We want to implement a retrieval engine only based on the computation of this fuzzy Galois connection. Another research area we want to explore is a combination of the direct and reverse implication, as suggested by the initial general logic framework proposed by Nie. As work in progress, we propose to use the proposed extension of the fuzzy Galois connection in text mining to extract fuzzy association rules between terms and perform query expansion using these fuzzy associations between terms.

Integrated Generic Association Rule Based Classifier

Associative classification is a supervised classification method. Many experimental studies have shown that associative classification is a promising approach. There are several associative classification approaches. However, the latter suffer from a major drawback: the huge number of the generated classification rules which takes efforts to select the best ones in order to construct the classifier. To overcome such drawback, we propose in this paper a new direct associative classification method called IGARC, an improvement of GARC approach, that extracts directly generic associative classification rules from a training set in order to reduce the number of associative classification rules without jeopardizing the classification accuracy. A detailed description of this method is presented, as well as the experimentation study on 12 benchmark data sets proving that IGARC is highly competitive in terms of accuracy in comparison with popular classification approaches.

Automatic handedness detection from off-line handwriting

In forensics, the handedness detection or the classification of writers into left or right-handed helps investigators focusing more on a certain category of suspects. However, only a few studies have been carried out in this field. Classification of handwriting into a demographic category is generally performed in two steps: feature extraction and classification. In this study, we propose a system which extract characterizing features from handwritings and use those features to perform the classification of handwritings with regards to handedness. Classification rates are reported on the QUWI dataset, reaching almost 70% for Left and right Handwriting.

Formal context coverage based on isolated labels: An efficient solution for text feature extraction

Different available data as images, texts, or database may be mapped into an equivalent or approximate binary relation. A text may be considered as a binary relation relating sentences to words, while a numerical table may be represented by a binary relation after using some scaling approach. A social network may be also represented by a formal context. The objective of this paper is to present an original approach for covering a binary relation by formal concepts based on isolated single or multiple properties, i.e., those belonging to only one concept. As a matter of fact, isolated properties are efficiently used for discriminating and labeling concepts. The latter are used for browsing in a corpora, or in a document by navigating through associated labels. By using fringe relations, the presented approach compared to those of the literature has the advantage of offering a relevant feature of a context by significant labels. Carried out experiments show the benefits of the introduced approach.

GARC: a new associative classification approach

Many studies in data mining have proposed a new classification approach called associative classification. According to several reports associative classification achieves higher classification accuracy than do traditional classification approaches. However, the associative classification suffers from a major drawback: it is based on the use of a very large number of classification rules; and consequently takes efforts to select the best ones in order to construct the classifier. To overcome such drawback, we propose a new associative classification method called Garc that exploits a generic basis of association rules in order to reduce the number of association rules without jeopardizing the classification accuracy. Moreover, Garc proposes a new selection criterion called score, allowing to ameliorate the selection of the best rules during classification. Carried out experiments on 12 benchmark data sets indicate that Garc is highly competitive in terms of accuracy in comparison with popular associative classification methods.

General learning approach for event extraction: Case of management change event

Starting from an ontology of a targeted financial domain corresponding to transaction, performance and management change news, relevant segments of text containing at least a domain keyword are extracted. The linguistic pattern of each segment is automatically generated to serve initially as a learning model. Each pattern is composed of named entities, keywords and articulation words. Some generic named entities like organizations, persons, locations, dates and grammatical annotations are generated by an automatic tool. During the learning step, each relevant segment is manually annotated with respect to the targeted entities (roles) structuring an event of the ontology. Information extraction is processed by associating a role with a specific entity. By alignment of generic entities to specific entities, some strings of a text are automatically annotated. An original learning approach is presented. Experiments with the management change event showed how recognition rates are improved by using different generalization tools.

Data Mining, Reasoning and Incremental Information Retrieval through Non Enlargeable Rectangular Relation Coverage

Association rules extraction from a binary relation as well as reasoning and information retrieval are generally based on the initial representation of the binary relation as an adjacency matrix. This presents some inconvenience in terms of space memory and knowledge organization. A coverage of a binary relation by a minimal number of non enlargeable rectangles generally reduces memory space consumption without any loss of information. It also has the advantage of organizing objects and attributes contained in the binary relation into a conceptual representation. In this paper, we propose new algorithms to extract association rules (i.e. data mining), conclusions from initial attributes (i.e. reasoning), as well as retrieving the total objects satisfying some initial attributes, by using only the minimal coverage. Finally we propose an incremental approximate algorithm to update a binary relation organized as a set of non enlargeable rectangles. Two main operations are mostly used during the organization process: First, separation of existing rectangles when we delete some pairs. Second, join of rectangles when common properties are discovered, after addition or removal of elements from a binary context. The objective is the minimization of the number of rectangles and the maximization of their structure. The article also raises the problems of equational modeling of the minimization criteria, as well as incrementally providing equations to maintain them.

Financial events detection by conceptual news categorization

In the scope of Financial Watch project, several targeted events have been required by contacted users in banking and investment domains. Financial news are classified with respect of the list of desired events. In this paper, a conceptual approach for indexing short English news in the financial domain is presented. By using a supervised original learning approach, a categorization method is proposed. Experimentation of the method on a sample of news showed that in almost all cases, occurrences of the right events in each document have been recognized, with respect to a corpus on the financial domain.