A associação é uma tarefa de mineração de dados que tem sido muito utilizada em problemas reais, porém o grande número de regras de associação que podem ser geradas dificulta a identificação de conhecimento interessante aos usuários. Para apoiar a identificação de regras interessantes podem ser utilizadas medidas de avaliação de conhecimento, que normalmente são classificadas como objetivas ou subjetivas. As medidas objetivas são mais gerais, mas podem não ser suficientes por não considerarem aspectos relacionados ao usuário ou ao domínio da aplicação. Por outro lado pode haver dificuldade em se obter a subjetividade do usuário necessária para o cálculo das medidas subjetivas. Diante desse contexto, neste trabalho é proposta uma metodologia para identificação de regras de associação interessantes que combina análises com medidas objetivas e subjetivas, visando aproveitar as vantagens de cada tipo e facilitar a participação do especialista. As medidas objetivas são utilizadas para selecionar algumas regras potencialmente interessantes para serem avaliadas por um especialista. As medidas subjetivas são calculadas utilizando essas regras com as avaliações do especialista. Essas medidas subjetivas então são utilizadas para auxiliar a identificação de regras interessantes de acordo com o conhecimento obtido durante a avaliação. Para viabilizar a aplicação dessa metodologia foi desenvolvido um módulo computacional de exploração de regras de associação com medidas subjetivas, denominado RulEE-SEAR. Utilizando esse módulo e outras ferramentas já existentes foi realizado um estudo de caso com uma base de dados real sobre qualidade de vida urbana. Nesse estudo de caso o processo de identificação de regras de associação interessantes foi realizado com especialista da área e verificou-se a viabilidade da metodologia proposta. / Association is a data mining task which has been applied in several real problems. However, due to the huge number of association rules that can be generated, it is hard for users to identify interesting knowledge. To assist users in finding interesting rules, evaluation measures can be used. Those measures are usually divided into objective and subjective. Objective measures are more general, but they can be insufficient because they do not consider user's and domain's features. On the other hand, getting users's knowledge and interest needed to calculate subjective measures can be a difficult task. In this context, a methodology to identify interesting association rules is proposed in this work. This methodology combines analysis with objective and subjective measures, aiming to use the advantages of each kind of measure and to make user's participation easier. Objective measures are used to select some potentially interesting rules for the user's evaluation. These rules and the evaluation are used to calculate subjective measures. Then, the subjective measures are used to assist the user in identifying interesting rules according to the knowledge obtained during the evaluation. To make the methodology use practicable, a computational module, named RulEE-SEAR, was developed to explore the association rules with subjective measures. Using this module and other existing tools, a case study was done. A urban life quality database was used and a specialist in this area participated in the interesting association rules identification. That case study showed that the methodology proposed is feasible.

The work gives overview of its three main topics: semantic web, information extraction and ontology learning. A method for identification relevant information on web pages is described and experimentally tested on pages of companies offering products and services. The method is based on analysis of a sample web pages and their position in the Open Directory catalogue. Furthermore, a modfication of association rules mining algorithm is proposed and experimentally tested. In addition to an identification of a relation between ontology concepts, it suggest possible naming of the relation.

Mineração de Dados é um processo de natureza iterativa e interativa responsável por identificar padrões em grandes conjuntos de dados, objetivando extrair conhecimento válido, útil e inovador a partir desses. Em Mineração de Dados, Regras de Associação é uma técnica que consiste na identificação de padrões intrínsecos ao conjunto de dados. Essa técnica tem despertado grande interesse nos pesquisadores de Mineração de Dados e nas organizações, entretanto, a mesma possui o inconveniente de gerar grande volume de conhecimento no formato de regras, dificultando a análise e interpretação dos resultados pelo usuário. Nesse contexto, este trabalho tem como objetivo principal generalizar e eliminar Regras de Associação não interessantes e/ou redundantes, facilitando, dessa maneira, a análise das regras obtidas com relação à compreensibilidade e tamanho do conjunto de regras. A generalização das Regras de Associação é realizada com o uso de taxonomias. Entre os principais resultados deste trabalho destacam-se a proposta e a implementação do algoritmo GART e do módulo computacional RulEE-GAR. O algoritmo GART (Generalization of Association Rules using Taxonomies - Generalização de Regras de Associação usando Taxonomias) utiliza taxonomias para generalizar Regras de Associação. Já o módulo RulEE-GAR, além de facilitar o uso do algoritmo GART durante a identificação de taxonomias e generalização de regras, provê funcionalidades para analisar as Regras de Associação generalizadas. Os experimentos realizados, neste trabalho, mostraram que o uso de taxonomias na generalização de Regras de Associação pode reduzir o volume de um conjunto de regras. / Data Mining refers to the process of finding patterns in large data sets. The Association Rules in Data Mining try to identify intrinsic behaviors of the data set. This has motivated researchers of Data Mining and organizations. However, the Association Rules have the inconvenient of generating a great amount of knowledge in the form of rules. This makes the analysis and interpretation of the results difficult for the user. Taking this into account, the main objective of this research is the generalization and elimination of non-interesting and/or redundant Association Rules. This facilite the analysis of the rules with respect to the compreensibility and the size of the rule set. The generalization is realized using taxonomies. The main results of this research are the proposal and the implementation of the algorithm GART and of the computational module RulEE-GAR. The algorithm GART (Generalization of Association Rules using Taxonomies) uses taxonomies to generalize Association Rules. The module RulEE-GAR facilitates the use of the algorithm GART in the identification of taxonomies and generalization of rules and provide functionalities to the analysis of the generalized Association Rules. The results of experiments showed that the employment of taxonomies in the generalization of Association Rules can reduce the size of a rule set.

Dentre as técnicas de mineração de dados encontra-se a associação, a qual identifica todas as associações intrínsecas contidas na base de dados. Entretanto, essa característica, vantajosa por um lado, faz com que um grande número de padrões seja gerado, sendo que muito deles, mesmo sendo estatisticamente aceitos, são triviais, falsos, ou irrelevantes à aplicação. Além disso, a técnica de associação tradicional gera padrões compostos apenas por itens contidos na base de dados, o que leva à extração, em geral, de um conhecimento muito específico. Essa especificidade dificulta a obtenção de uma visão geral do domínio pelos usuários finais, que visam a utilização/exploração de conhecimentos úteis e compreensíveis. Assim, o pós-processamento das regras descobertas se torna um importante tópico, uma vez que há a necessidade de se validar as regras obtidas. Diante do exposto, este trabalho apresenta uma abordagem de pós-processamento de regras de associação que utiliza conhecimento de domínio, expresso via taxonomias, para obter um conjunto de regras de associação generalizadas compacto e representativo. Além disso, a fim de avaliar a representatividade de padrões generalizados, é apresentado também neste trabalho um estudo referente à utilização de medidas de interesse objetivas quando aplicadas a regras de associação generalizadas. Nesse estudo, a semântica da generalização é levada em consideração, já que cada uma delas fornece uma visão distinta do domínio. Como resultados desta tese, foi possível observar que: um conjunto de regras de associação pode ser compactado na presença de um conjunto de taxonomias; para cada uma das semânticas de generalização existe um conjunto de medidas mais apropriado para ser utilizado na avaliação de regras generalizadas / The association technique, one of the data mining techniques, identifies all the intrinsic associations in database. This characteristic, which can be advantageous on the one hand, generates a large number of patterns. Many of these patterns, even statistically accepted, are trivial, spurious, or irrelevant to the application. In addition, the association technique generates patterns composed only by items in database, which in general implies a very specific knowledge. This specificity makes it difficult to obtain a general view of the domain by the final users, who aims the utilization/exploration of useful and comprehensible knowledge . Thus, the post-processing of the discovered rules becomes an important topic, since it is necessary to validate the obtained rules. In this context, this work presents an approach for post-processing association rules that uses domain knowledge, expressed by taxonomies, to obtain a reduced and representative generalized association rule set. In addition, in order to evaluate the representativeness of generalized patterns, a study referent to the use of objective interest measures when applied to generalized association rules is presented. In this study, the generalization semantics is considered, since each semantic provides a distinct view of the domain. As results of this thesis, it was possible to observe that: an association rule set can be compacted with a taxonomy set; for each generalization semantic there is a measure set that is more appropriate to be used in the generalized rules evaluation

With the growth of geo-referenced data and the sophistication and complexity of spatial databases, data mining and knowledge discovery techniques become essential tools for successful analysis of large spatial datasets. Spatial clustering is fundamental and central to geographical data mining. It partitions a dataset into smaller homogeneous groups due to spatial proximity. Resulting groups represent geographically interesting patterns of concentrations for which further investigations should be undertaken to find possible causal factors. In this thesis, we propose a spatial-dominant generalization approach that mines multivariate causal associations among geographical data layers using clustering analysis. First, we propose a generic framework of multi-purpose exploratory spatial clustering in the form of the Template-Method Pattern. Based on an object-oriented framework, we design and implement an automatic multi-purpose exploratory spatial clustering tool. The first instance of this framework uses the Delaunay diagram as an underlying proximity graph. Our spatial clustering incorporates the peculiar characteristics of spatial data that make space special. Thus, our method is able to identify high-quality spatial clusters including clusters of arbitrary shapes, clusters of heterogeneous densities, clusters of different sizes, closely located high-density clusters, clusters connected by multiple chains, sparse clusters near to high-density clusters and clusters containing clusters within O(n log n) time. It derives values for parameters from data and thus maximizes user-friendliness. Therefore, our approach minimizes user-oriented bias and constraints that hinder exploratory data analysis and geographical data mining. Sheer volume of spatial data stored in spatial databases is not the only concern. The heterogeneity of datasets is a common issue in data-rich environments, but left open by exploratory tools. Our spatial clustering extends to the Minkowski metric in the absence or presence of obstacles to deal with situations where interactions between spatial objects are not adequately modeled by the Euclidean distance. The genericity is such that our clustering methodology extends to various spatial proximity graphs beyond the default Delaunay diagram. We also investigate an extension of our clustering to higher-dimensional datasets that robustly identify higher-dimensional clusters within O(n log n) time. The versatility of our clustering is further illustrated with its deployment to multi-level clustering. We develop a multi-level clustering method that reveals hierarchical structures hidden in complex datasets within O(n log n) time. We also introduce weighted dendrograms to effectively visualize the cluster hierarchies. Interpretability and usability of clustering results are of great importance. We propose an automatic pattern spotter that reveals high level description of clusters. We develop an effective and efficient cluster polygonization process towards mining causal associations. It automatically approximates shapes of clusters and robustly reveals asymmetric causal associations among data layers. Since it does not require domain-specific concept hierarchies, its applicability is enhanced. / PhD Doctorate

Data mining techniques have been widely applied to extract desirable knowledge from existing databases for specific purposes. In real-world applications, a database usually involves the time periods when transactions occurred and exhibition periods of items, in addition to the items bought in the transactions. To handle this kind of data, temporal data mining techniques are thus proposed to find temporal association rules from a database with time. Most of the existing studies only consider different item lifespans to find general temporal association rules, and this may neglect some useful information. For example, while an item within the whole exhibition period may not be a frequent one, it may be frequent within part of this time. To deal with this, the concept of a hierarchy of time is thus applied to temporal data mining along with suitable time granules, as defined by users. In this thesis, we thus handle the problem of mining temporal association rules with a hierarchy of time granules from a temporal database, and also propose three novel mining algorithms for different item lifespan definitions. In the first definition, the lifespan of an item in a time granule is calculated from the first appearance time to the end time in the time granule. In the second definition, the lifespan of an item in a time granule is evaluated from the publication time of the item to the end time in the time granule. Finally, in the third definition, the lifespan of an item in a time granule is measured by its entire exhibition period. The experimental results on a simulation dataset show the performance of the three proposed algorithms under different item lifespan definitions, and compare the mined temporal association rules with and without consideration of the hierarchy of time granules under different parameter settings.

Over the past few years, mining data located in heterogeneous and geographically distributed sites have been designated as one of the key important issues. Loading distributed data into centralized location for mining interesting rules is not a good approach. This is because it violates common issues such as data privacy and it imposes network overheads. The situation becomes worse when the network has limited bandwidth which is the case in most of the real time systems. This has prompted the need for intelligent data analysis to discover the hidden information in these huge amounts of distributed databases. In this research, we present an incremental approach for building an efficient Multi-Agent based algorithm for mining real world databases in geographically distributed sites. First, we propose the Distributed Multi-Agent Association Rules algorithm (DMAAR) to minimize the all-to-all broadcasting between distributed sites. Analytical calculations show that DMAAR reduces the algorithm complexity and minimizes the message communication cost. The proposed Multi-Agent based algorithm complies with the Foundation for Intelligent Physical Agents (FIPA), which is considered as the global standards in communication between agents, thus, enabling the proposed algorithm agents to cooperate with other standard agents. Second, the BitTable Multi-Agent Association Rules algorithm (BMAAR) is proposed. BMAAR includes an efficient BitTable data structure which helps in compressing the database thus can easily fit into the memory of the local sites. It also includes two BitWise AND/OR operations for quick candidate itemsets generation and support counting. Moreover, the algorithm includes three transaction trimming techniques to reduce the size of the mined data. Third, we propose the Pruning Multi-Agent Association Rules algorithm (PMAAR) which includes three candidate itemsets pruning techniques for reducing the large number of generated candidate itemsets, consequently, reducing the total time for the mining process. The proposed PMAAR algorithm has been compared with existing Association Rules algorithms against different benchmark datasets and has proved to have better performance and execution time. Moreover, PMAAR has been implemented on real world distributed medical databases obtained from more than one hospital in Egypt to discover the hidden Association Rules in patients' records to demonstrate the merits and capabilities of the proposed model further. Medical data was anonymously obtained without the patients' personal details. The analysis helped to identify the existence or the absence of the disease based on minimum number of effective examinations and tests. Thus, the proposed algorithm can help in providing accurate medical decisions based on cost effective treatments, improving the medical service for the patients, reducing the real time response for the health system and improving the quality of clinical decision making.

Η παρούσα μεταπτυχιακή εργασία έχει στόχο τη μελέτη προβλημάτων «κρυμμένης γνώσης» από συστήματα και εφαρμογές ηλεκτρονικού εμπορίου (e-commerce) και ηλεκτρονικής μάθησης (e-learning) με κύριο στόχο τη βελτίωση της ποιότητας και της απόδοσης των παρεχόμενων υπηρεσιών προς τους τελικούς χρήστες. Στο πρώτο κεφάλαιο παρουσιάζεται ένα σενάριο για σημασιολογικά εξατομικευμένο e-learning. Ο προτεινόμενος αλγόριθμος βασίζεται σε μια οντολογία (ontology) η οποία βοηθά στη δόμηση και στη διαχείριση του περιεχομένου που σχετίζεται με μια δεδομένη σειρά μαθημάτων, ένα μάθημα ή ένα θεματικό. Η διαδικασία χωρίζεται σε δύο στάδια: στο offline στάδιο το οποίο περιλαμβάνει τις ενέργειες προετοιμασίας των δεδομένων, δημιουργίας της οντολογίας και εξόρυξης από δεδομένα χρήσης (usage mining) και στο online στάδιο το οποίο περιλαμβάνει την εξαγωγή των εξατομικευμένων συστάσεων. Το προτεινόμενο σύστημα σε πρώτη φάση βρίσκει ένα αρχικό σύνολο συστάσεων βασισμένο στην οντολογία του πεδίου και στη συνέχεια χρησιμοποιεί τα frequent itemsets (συχνά εμφανιζόμενα σύνολα στοιχείων) για να το εμπλουτίσει, λαμβάνοντας υπόψη την πλοήγηση άλλων παρόμοιων χρηστών (similar users). Με τον τρόπο αυτό, μειώνεται ο χρόνος που απαιτείται για την ανάλυση όλων των frequent itemsets και των κανόνων συσχέτισης. Η ανάλυση εστιάζεται μόνο σε εκείνα τα σύνολα που προέρχονται από το συνδυασμό της ενεργούς συνόδου (current session) του χρήστη και των συστάσεων της οντολογίας. Αν και η εξατομίκευση απαιτεί αρκετά βήματα επεξεργασίας και ανάλυσης, στη συγκεκριμένη προσέγγιση το εμπόδιο αυτό αποφεύγεται με την εκτέλεση σημαντικού μέρους της διαδικασίας offline. Στο δεύτερο κεφάλαιο μελετάται το πρόβλημα της παραγωγής προτάσεων σε μια εφαρμογή e-commerce. Η προτεινόμενη υβριδική προσέγγιση στοχεύει στην παραγωγή αποτελεσματικών συστάσεων για τους πελάτες ενός online καταστήματος που ενοικιάζει κινηματογραφικές ταινίες. Η γνώση για τους πελάτες και τα προϊόντα προκύπτει από τα δεδομένα χρήσης και τη δομή της οντολογίας σε συνδυασμό με τις εκτιμήσεις-βαθμολογίες των πελατών για τις ταινίες καθώς και την εφαρμογή τεχνικών ταιριάσματος «όμοιων» πελατών. Όταν ένα ή περισσότερα κριτήρια ταιριάσματος ικανοποιούνται, τότε άλλες ταινίες μπορούν να προσδιοριστούν σύμφωνα με το οντολογικό σχήμα που έχουν παρόμοια χαρακτηριστικά με αυτές που ο πελάτης έχει ήδη νοικιάσει. Στην περίπτωση ενός νέου πελάτη όπου το ιστορικό του είναι κενό, αναλύονται πληροφορίες από την αίτηση εγγραφής του ώστε να ταξινομηθεί σε μια συγκεκριμένη κλάση πελατών και να παραχθούν προτάσεις με βάση το οντολογικό σχήμα. Αυτή η ενοποίηση παρέχει πρόσθετη γνώση για τις προτιμήσεις των πελατών και επιτρέπει την παραγωγή επιτυχημένων συστάσεων. Ακόμη και στην περίπτωση του «cold-start problem» όπου δεν είναι διαθέσιμη αρχική πληροφορία για τη συμπεριφορά του πελάτη, η προσέγγιση προβαίνει σε σχετικές συστάσεις. Στο τρίτο κεφάλαιο παρουσιάζεται μία νέα προσέγγιση στο πρόβλημα της δημιουργίας συστάσεων. Οι προηγούμενες προσεγγίσεις δεν λαμβάνουν υπόψη τους τη σειρά με την οποία ο χρήστης προσπελαύνει τα δεδομένα, είτε πρόκειται για e-learning είτε πρόκειται για e-commerce δεδομένα. Στο κεφάλαιο αυτό προτείνεται μία τεχνική η οποία λαμβάνει υπόψη τη σειρά με την οποία ο χρήστης προσπελαύνει τα δεδομένα (ordering). Πιο συγκεκριμένα μελετάται η τεχνική αυτή σε e-commerce συστήματα και καλάθια αγορών. Παρουσιάζεται και αναλύεται η υλοποίηση του προτεινόμενου αλγορίθμου. Επιπλέον γίνεται αξιολόγηση των αποτελεσμάτων του αλγορίθμου σε testing input data τα οποία και δείχνουν την ποιότητα των παραγόμενων συστάσεων. / -

The main goal of these thesis is to compare association rules finding algorithms and to indicate the usability of finding association rules in business area. In order to achieve this goal, the theoretical analysis of three algorithms is done: 1. The Apriori algorithm – the most well known association rule algorithm – based on the property: “Any subset of a large itemset must be large”. This algorithm assumes that the database is memory-resident. The maximum number of database scans is one more than the cardinality of the largest large itemset. 2. The Sampling algorithm deals with the database sample prior the full database scan. The database sample is drawn such that it can be memory-resident. The Sampling algorithm reduces the number of database scans to one in the best case and two in the worst case. 3. The Partitioning algorithm divides database into partitions and bases on the property: “A large itemset must be large in at least one of the partitions”. This algorithm reduces the number of database scans to two and divides the database into partitions such that each partition can be placed into main memory. There are created programs for all three algorithms plus the program for the full set of itemsets algorithm. Programs are created in C++ language. In order to achieve topmost performance, the GUI is missed. Nine test data sets are created to compare the algorithms. Six of them contains real life data from telecommunications business area. Datasets varies from the... [to full text]

This thesis is concerned with the merging of two active research domains: Knowledge Discovery in Databases (KDD), more precisely the Association Rule Mining technique, and Knowledge Engineering (KE) with a main interest in knowledge representation languages developed around the Semantic Web. In Data Mining, the usefulness of association rule technique is strongly limited by the huge amount and the low quality of delivered rules. Experiments show that rules become almost impossible to use when their number exceeds 100. At the same time, nuggets are often represented by those rare (low support) unexpected association rules which are surprising to the user. Unfortunately, the lower the support is, the larger the volume of rules becomes. Thus, it is crucial to help the decision maker with an efficient technique to reduce the number of rules. To overcome this drawback, several methods have been proposed in the literature such as itemset concise representations, redundancy reduction, filtering, ranking and post-processing. Even though rule interestingness strongly depends on user knowledge and goals, most of the existing methods are generally based on data structure. For instance, if the user looks for unexpected rules, all the already known rules should be pruned. Or, if the user wants to focus on specific family of rules, only this subset of rules should be selected. In this context, we address two main issues: the integration of user knowledge in the discovery process and the interactivity with the user. The first issue requires defining an adapted formalism to express user knowledge with accuracy and flexibility such as ontologies in the Semantic Web. Second, the interactivity with the user allows a more iterative mining process where the user can successively test different hypotheses or preferences and focus on interesting rules. The main contributions of this work can be summarized as follows: (i) A model to represent user knowledge. First, we propose a new rule-like formalism, called Rule Schema, which allows the user to define his/her expectations regarding the rules through ontology concepts. Second, ontologies allow the user to express his/her domain knowledge by means of a high semantic model. Last, the user can choose among a set of Operators for interactive processing the one to be applied over each Rule Schema (i.e. pruning, conforming, unexpectedness, . . . ). (ii) A new post-processing approach, called ARIPSO (Association Rule Interactive Post-processing using rule Schemas and Ontologies), which helps the user to reduce the volume of the discovered rules and to improve their quality. It consists in an interactive process integrating user knowledge and expectations by means of the proposed model. At each step of ARIPSO, the interactive loop allows the user to change the provided information and to reiterate the post-processing phase which produces new results. (iii) The implementation in post-processing of the proposed approach. The developed tool is complete and operational, and it implements all the functionalities described in the approach. Also, it makes the connection between different elements like the set of rules and rule schemas stored in PMML/XML files, and the ontologies stored in OWL files and inferred by the Pellet reasoner. (iv) An adapted implementation without post-processing, called ARLIUS (Association Rule Local mining Interactive Using rule Schemas), consisting in an interactive local mining process guided by the user. It allows the user to focus on interesting rules without the necessity to extract all of them, and without minimum support limit. In this way, the user may explore the rule space incrementally, a small amount at each step, starting from his/her own expectations and discovering their related rules. (v) The experimental study analyzing the approach efficiency and the discovered rule quality. For this purpose, we used a real-life and large questionnaire database concerning customer satisfaction. For ARIPSO, the experimentation was carried out in complete cooperation with the domain expert. For different scenarios, from an input set of nearly 400 thousand association rules, ARIPSO filtered between 3 and 200 rules validated by the expert. Clearly, ARIPSO allows the user to significantly and efficiently reduce the input rule set. For ARLIUS, we experimented different scenarios over the same questionnaire database and we obtained reduced sets of rules (less than 100) with very low support.