Geetha Mary Amalanathan

Ce livre propose une introduction compl te au data mining, couvrant les concepts fondamentaux, les techniques essentielles et les applications pratiques dans diff rents domaines. De l'exploration des m thodes de pr traitement des donn es la plong e dans les algorithmes avanc s tels que les arbres de d cision, les r seaux neuronaux et le clustering, chaque chapitre offre des aper us th oriques compl t s par des exemples et des exercices du monde r el. Que vous soyez tudiant ou professionnel, ce livre vous fournit les outils n cessaires pour extraire des informations pr cieuses des donn es, vous permettant ainsi de prendre des d cisions clair es et de stimuler l'innovation dans votre domaine.

Mit dem Wachstum des digitalen Zeitalters sind Daten in gro em Umfang verf gbar, so dass das Abrufen von Wissen aus diesen Daten durch Data-Mining-Algorithmen erfolgt. Unter den verschiedenen Data-Mining-Algorithmen ist die Erkennung von Ausrei ern von entscheidender Bedeutung, da ihr Auftreten die Effizienz des Systems beeintr chtigt. Die meisten Forschungsarbeiten beschr nken sich auf die Erkennung von Ausrei ern in einem einzigen Universum mit einer einzigen Granulation f r numerische oder kategoriale Daten. Die vorhandenen Algorithmen zur Erkennung von Ausrei ern durch maschinelles Lernen funktionieren gut bei quantitativen Daten, lassen sich aber nicht direkt auf qualitative, vage und ungenaue Daten anwenden, was zu ineffektiven Ergebnissen f hrt. In der realen Welt gibt es auch mehrdeutige, unsichere, unvollst ndige und unbestimmte Informationen . Diese Probleme werden in dieser Forschungsarbeit mit Hilfe der Theorie der groben Mengen, der intuitionistischen Fuzzy-Mengen und der neutrosophischen Mengen behandelt . Die vorgeschlagene Methode zur Erkennung von Ausrei ern auf der Grundlage von grober Entropie und gewichteter Dichte wurde entwickelt, um Ausrei er in verschiedenen Informationssystemen zu erkennen . Der gewichtete Dichtewert f r jedes Objekt und jedes Attribut wurde bestimmt, um Ausrei er zu erkennen. So wird ein echtes Objekt niemals als Ausrei erbehandelt.

Avec la croissance de l' re num rique, les donn es sont largement disponibles, de sorte que la recherche de connaissances partir de ces donn es est effectu e par des algorithmes d'exploration de donn es. Parmi les diff rents algorithmes d'exploration de donn es, la d tection des valeurs aberrantes est cruciale, car leur pr sence d grade l'efficacit du syst me. La majorit des recherches se sont limit es la d tection des valeurs aberrantes dans un seul univers avec une seule granulation pour les donn es num riques ou cat gorielles. Les algorithmes existants de d tection des valeurs aberrantes par apprentissage automatique fonctionnent bien pour les donn es quantitatives, mais ils ne sont pas directement appliqu s aux donn es qualitatives, vagues et impr cises, ce qui produit des r sultats inefficaces. Il existe galement des informations ambigu s, incertaines, incompl tes et ind termin es qui persistent dans le monde r el. Ces probl mes sont trait s dans ce travail de recherche l'aide de la th orie des ensembles rugueux, des ensembles flous intuitionnistes et des ensembles neutrosophiques. La m thodologie propos e, la m thode de d tection des valeurs aberrantes densit pond r e bas e sur l'entropie grossi re, a t con ue pour d tecter les valeurs aberrantes dans divers syst mes d'information.

Con la crescita dell'era digitale, i dati sono ampiamente disponibili e il recupero della conoscenza da questi dati viene effettuato con algoritmi di data mining. Tra i vari algoritmi di data mining, l'individuazione degli outlier fondamentale, poich la loro presenza degrada l'efficienza del sistema. La maggior parte della ricerca si limitata a rilevare gli outlier in un singolo universo con una singola granulazione per dati numerici o categorici. Gli algoritmi di rilevamento degli outlier di apprendimento automatico esistenti funzionano bene per i dati quantitativi, ma non sono direttamente applicabili a dati qualitativi, vaghi e imprecisi, il che produce risultati inefficaci. Nel mondo reale esistono anche informazioni ambigue, incerte, incomplete e indeterminate . Questi problemi vengono affrontati in questo lavoro di ricerca utilizzando la teoria degli insiemi grezzi, gli insiemi fuzzy intuitivisti e neutrosofici. La metodologia proposta, basata sull'entropia grezza e sulla densit ponderata, stata progettata per rilevare gli outlier in vari sistemi informativi. Il valore della densit ponderata per ogni oggetto e attributo stato determinato per rilevare gli outlier. In questo modo, un oggetto vero non sar mai trattato come un outlier.

Com o crescimento da era digital, os dados est o amplamente dispon veis, pelo que a recupera o de conhecimentos a partir desses dados efectuada atrav s de algoritmos de extra o de dados. Entre os v rios algoritmos de extra o de dados, a dete o de valores an malos crucial, uma vez que a sua ocorr ncia degrada a efici ncia do sistema. A maior parte da investiga o limitou-se dete o de valores at picos num nico universo com uma nica granula o para dados num ricos ou categ ricos. Os algoritmos de dete o de valores at picos de aprendizagem autom tica existentes funcionam bem para dados quantitativos, mas n o s o diretamente aplicados a dados qualitativos, vagos e imprecisos, o que produz resultados ineficazes. H tamb m informa es amb guas, incertas, incompletas e indeterminadas que persistem no mundo real. Estes problemas s o tratados neste trabalho de investiga o utilizando a teoria dos conjuntos aproximados, os conjuntos difusos intuicionistas e os conjuntos neutros ficos. A metodologia proposta, baseada na entropia grosseira e no m todo de dete o de valores aberrantes de densidade ponderada, foi concebida para detetar valores aberrantes em v rios sistemas de informa o. O valor da densidade ponderada para cada objeto e atributo foi determinado para detetar anomalias. Assim, um objeto verdadeiro nunca ser tratado como um outlier.

This book provides a comprehensive introduction to data mining, covering fundamental concepts, essential techniques and practical applications across various domains. From exploring data preprocessing methods to diving into advanced algorithms like decision trees, neural networks and clustering, each chapter offers theoretical insights complemented by real-world examples and exercises. Whether you're a student or a professional, this book equips you with the tools to extract valuable insights from data, empowering you to make informed decisions and drive innovation in your field.

In this digital transformation era, data is growing exponentially due to substantial online transactions and the internet of things, etc. As data is growing, it needs efficient data management and scalable storage mechanisms to deal with the vast data that is generated. Now plenty of tools, technologies, and frameworks are available that can handle this kind of Big Data scenario with distributed systems that can scale horizontally on the cloud platforms. Master Data Management (MDM) is a technology-enabled discipline in which the information technologies team works closely with business teams to ensure data accountability to achieve uniformity, accuracy, consistency, a single source of truth, etc. Current MDM systems cannot address quality-related issues such as reliable, real-time data synchronization, deduplication, entity resolution, etc., due to multi-folded data growth. The following four major gaps from the above quality barriers can be drawn as gaps in the research and have been addressed in this book. - Unclear master data definition- Lack of responsibilities in data maintenance- Inaccurate data matching- Immature data deduplication process.

With the growth of the digital era, data is largely available, so knowledge retrieval from those data is done by data mining algorithms. Among various data mining algorithms, finding outliers is crucial as their occurrence degrades system efficiency. The majority of the research was limited to detecting outliers in a single universe with a single granulation for numerical or categorical data. The existing machine learning outlier detection algorithms work well for quantitative data but they are not directly applied to qualitative, vague and imprecise data which produces ineffective results. There is also ambiguous, uncertain, incomplete, and indeterminate information that persists in this real world. These problems are handled in this research work using rough set theory, intuitionistic fuzzy, and neutrosophic sets. The proposed methodology rough entropy based weighted density outlier detection method has been designed to detect outliers for various information systems. The weighted density value for each object and attribute has been determined to detect outliers. So a true object will never be treated as an outlier.