11 |
Extração de conhecimento de redes neurais artificiais. / Knowledge extraction from artificial neural networks.Edmar Martineli 20 August 1999 (has links)
Este trabalho descreve experimentos realizados com Redes Neurais Artificiais e algoritmos de aprendizado simbólico. Também são investigados dois algoritmos de extração de conhecimento de Redes Neurais Artificiais. Esses experimentos são realizados com três bases de dados com o objetivo de comparar os desempenhos obtidos. As bases de dados utilizadas neste trabalho são: dados de falência de bancos brasileiros, dados do jogo da velha e dados de análise de crédito. São aplicadas sobre os dados três técnicas para melhoria de seus desempenhos. Essas técnicas são: partição pela menor classe, acréscimo de ruído nos exemplos da menor classe e seleção de atributos mais relevantes. Além da análise do desempenho obtido, também é feita uma análise da dificuldade de compreensão do conhecimento extraído por cada método em cada uma das bases de dados. / This work describes experiments carried out witch Artificial Neural Networks and symbolic learning algorithms. Two algorithms for knowledge extraction from Artificial Neural Networks are also investigates. This experiments are performed whit three data set with the objective of compare the performance obtained. The data set used in this work are: Brazilians banks bankruptcy data set, tic-tac-toe data set and credit analysis data set. Three techniques for data set performance improvements are investigates. These techniques are: partition for the smallest class, noise increment in the examples of the smallest class and selection of more important attributes. Besides the analysis of the performance obtained, an analysis of the understanding difficulty of the knowledge extracted by each method in each data bases is made.
|
12 |
Large-Scale Multilingual Knowledge Extraction, Publishing and Quality Assessment: The case of DBpediaKontokostas, Dimitrios 04 September 2018 (has links)
No description available.
|
13 |
Fouille de données par extraction de motifs graduels : contextualisation et enrichissement / Data mining based on gradual itemsets extraction : contextualization and enrichmentOudni, Amal 09 July 2014 (has links)
Les travaux de cette thèse s'inscrivent dans le cadre de l'extraction de connaissances et de la fouille de données appliquée à des bases de données numériques ou floues afin d'extraire des résumés linguistiques sous la forme de motifs graduels exprimant des corrélations de co-variations des valeurs des attributs, de la forme « plus la température augmente, plus la pression augmente ». Notre objectif est de les contextualiser et de les enrichir en proposant différents types de compléments d'information afin d'augmenter leur qualité et leur apporter une meilleure interprétation. Nous proposons quatre formes de nouveaux motifs : nous avons tout d'abord étudié les motifs dits « renforcés », qui effectuent, dans le cas de données floues, une contextualisation par intégration d'attributs complémentaires, ajoutant des clauses introduites linguistiquement par l'expression « d'autant plus que ». Ils peuvent être illustrés par l'exemple « plus la température diminue, plus le volume de l'air diminue, d'autant plus que sa densité augmente ». Ce renforcement est interprété comme validité accrue des motifs graduels. Nous nous sommes également intéressées à la transposition de la notion de renforcement aux règles d'association classiques en discutant de leurs interprétations possibles et nous montrons leur apport limité. Nous proposons ensuite de traiter le problème des motifs graduels contradictoires rencontré par exemple lors de l'extraction simultanée des deux motifs « plus la température augmente, plus l'humidité augmente » et « plus la température augmente, plus l'humidité diminue ». Pour gérer ces contradictions, nous proposons une définition contrainte du support d'un motif graduel, qui, en particulier, ne dépend pas uniquement du motif considéré, mais aussi de ses contradicteurs potentiels. Nous proposons également deux méthodes d'extraction, respectivement basées sur un filtrage a posteriori et sur l'intégration de la contrainte du nouveau support dans le processus de génération. Nous introduisons également les motifs graduels caractérisés, définis par l'ajout d'une clause linguistiquement introduite par l'expression « surtout si » comme par exemple « plus la température diminue, plus l'humidité diminue, surtout si la température varie dans [0, 10] °C » : la clause additionnelle précise des plages de valeurs sur lesquelles la validité des motifs est accrue. Nous formalisons la qualité de cet enrichissement comme un compromis entre deux contraintes imposées à l'intervalle identifié, portant sur sa taille et sa validité, ainsi qu'une extension tenant compte de la densité des données.Nous proposons une méthode d'extraction automatique basée sur des outils de morphologie mathématique et la définition d'un filtre approprié et transcription. / This thesis's works belongs to the framework of knowledge extraction and data mining applied to numerical or fuzzy data in order to extract linguistic summaries in the form of gradual itemsets: the latter express correlation between attribute values of the form « the more the temperature increases, the more the pressure increases ». Our goal is to contextualize and enrich these gradual itemsets by proposing different types of additional information so as to increase their quality and provide a better interpretation. We propose four types of new itemsets: first of all, reinforced gradual itemsets, in the case of fuzzy data, perform a contextualization by integrating additional attributes linguistically introduced by the expression « all the more ». They can be illustrated by the example « the more the temperature decreases, the more the volume of air decreases, all the more its density increases ». Reinforcement is interpreted as increased validity of the gradual itemset. In addition, we study the extension of the concept of reinforcement to association rules, discussing their possible interpretations and showing their limited contribution. We then propose to process the contradictory itemsets that arise for example in the case of simultaneous extraction of « the more the temperature increases, the more the humidity increases » and « the more the temperature increases, the less the humidity decreases ». To manage these contradictions, we define a constrained variant of the gradual itemset support, which, in particular, does not only depend on the considered itemset, but also on its potential contradictors. We also propose two extraction methods: the first one consists in filtering, after all itemsets have been generated, and the second one integrates the filtering process within the generation step. We introduce characterized gradual itemsets, defined by adding a clause linguistically introduced by the expression « especially if » that can be illustrated by a sentence such as « the more the temperature decreases, the more the humidity decreases, especially if the temperature varies in [0, 10] °C »: the additional clause precise value ranges on which the validity of the itemset is increased. We formalize the quality of this enrichment as a trade-off between two constraints imposed to identified interval, namely a high validity and a high size, as well as an extension taking into account the data density. We propose a method to automatically extract characterized gradual based on appropriate mathematical morphology tools and the definition of an appropriate filter and transcription.
|
14 |
Investigation into Regression Analysis of Multivariate Additional Value and Missing Value Data Models Using Artificial Neural Networks and Imputation TechniquesJagirdar, Suresh 01 October 2008 (has links)
No description available.
|
15 |
Knowledge Graph Extension by Entity Type RecognitionShi, Daqian 23 April 2024 (has links)
Knowledge graphs have emerged as a sophisticated advancement and refinement of semantic networks, and their deployment is one of the critical methodologies in contemporary artificial intelligence. The construction of knowledge graphs is a multifaceted process involving various techniques, where researchers aim to extract the knowledge from existing resources for the construction since building from scratch entails significant labor and time costs. However, due to the pervasive issue of heterogeneity, the description diversity across different knowledge graphs can lead to mismatches between concepts, thereby impacting the efficacy of knowledge extraction. This Ph.D. study focuses on automatic knowledge graph extension, i.e., properly extending the reference knowledge graph by extracting and integrating concepts from one or more candidate knowledge graphs. We propose a novel knowledge graph extension framework based on entity type recognition. The framework aims to achieve high-quality knowledge extraction by aligning the schemas and entities across different knowledge graphs, thereby enhancing the performance of the extension. This paper elucidates three major contributions: (i) we propose an entity type recognition method exploiting machine learning and property-based similarities to enhance knowledge extraction; (ii) we introduce a set of assessment metrics to validate the quality of the extended knowledge graphs; (iii) we develop a platform for knowledge graph acquisition, management, and extension to benefit knowledge engineers practically. Our evaluation comprehensively demonstrated the feasibility and effectiveness of the proposed extension framework and its functionalities through quantitative experiments and case studies.
|
16 |
運用文字探勘技術輔助建構法律條文之語意網路-以公司法為例張露友 Unknown Date (has links)
本論文運用文字探勘相關技術,嘗試自動計算法條間的相似度,輔助專家從公司法眾多法條中整理出規則,建立法條之間的關聯,使整個法典並不是獨立的法條條號與法條內容的集合,而是在法條之間透過語意的方式連結成網路,並從分析與解釋關聯的過程中,探討文字探勘技術運用於法律條文上所遭受之困難及限制,以供後續欲從事相關研究之參考。
本論文的研究結果,從積極面來看,除了可以建立如何運用文字探勘於輔助法律知識擷取的方法之外,另一方面,從消極面來看,倘若研究結果顯示,文字探勘技術並不完全適用於法律條文的知識擷取上,那麼對於從事類似研究的專業人員而言,本論文所提出的結論與建議,亦可作為改善相關技術的重要參考。 / This thesis tries to use text mining technique to calculate, compare and analyze the correlation of legal codes. And based on the well-known defined legal concept and knowledge, it also tries to help explain and evaluate the relations above using the result of automatic calculation. Furthermore, this thesis also wishes to contribute on how to apply information technology effectively onto legal knowledge domain.
If the research reveals the positive result, it could be used for knowledge build-up on how to utilize text mining technology onto legal domain. However, if the study shows that text mining doesn’t apparently apply to knowledge extracting of legal domain, then the conclusion and suggestion from this thesis could also be regarded as a important reference to other professionals in the similar research fields.
|
17 |
Sociální sítě a dobývání znalostí / Social networks and data miningZvirinský, Peter January 2014 (has links)
Recent data mining methods represent modern approaches capable of analyzing large amounts of data and extracting meaningful and potentially useful information from it. In this work, we discuss all the essential steps of the data mining process - including data preparation, storage, cleaning, data analysis as well as visualization of the obtained results. In particular, this work is focused on the data available publicly from the Insolvency Register of the Czech Republic, that comprises all insolvency proceedings commenced after 1. January 2008 in the Czech Republic. With regard to the considered type of data, several data mining methods have been discussed, implemented, tested and evaluated. Among others, the studied techniques include Market Basket Analysis, Bayesian networks and social network analysis. The obtained results reveal several social patterns common in the current Czech society.
|
18 |
Knowledge technologies process and cultures : improving information and knowledge sharing at the Amateur Swimming Association (ASA)Onojeharho, Ejovwoke January 2015 (has links)
Over the last few years the ASA determined KM as a priority to assist with reducing knowledge loss, realising information assets and reducing work duplication by attempting to implement IKM tools and strategies. This research employed a pragmatic viewpoint, using a mix of both quantitative and qualitative methods to check reliability, to ensure validity while undertaking the task of implementing the IKM tools. Using a case study strategy and action research was justified, as to be pragmatic the researcher needed to understand the extent of the problem within a specified context. The research discussed in this thesis, provides a new framework for implementing KM tools; focusing on the NSO category, which the case study organisation falls into. The literature agrees enlisting influential members onto the project is vital for success; however, the findings suggested that success was not only tied to this buy-in alone, but also to the organisation s ability to retain these members for the duration of the project. The research proposed the use of a newly developed tool within the new framework, as an approach to reduce the time it takes to undertake traditional social network analysis of the organisation, as it became clear that there was a need for a method of producing updated results of the SNA, which would span the length of long projects within organisations with significantly high staff turn-over rates. Privacy was given as a factor to consider the in literature; however, the findings from this study indicated that a majority of the participants were comfortable with the system. Email knowledge extraction, and email social network systems are not new concepts, however this research presents EKESNA; a novel tool that combines both concepts in a way that allows for the continuous discovery, visualisation, and analysis of knowledge networks around specified topics of interest within an organisation; linking conversations to specific expert knowledge. EKESNA s continuous discovery of the organisation s knowledge network affords members up-to-date data to inform business process reengineering. This is a potentially ground breaking new tool that has the possibility of transforming the KM landscape in NSOs as well as a whole range of other kinds of enterprises.
|
19 |
Contribution to study and implementation of a bio-inspired perception system based on visual and auditory attention / Contribution à l’étude et à la mise en œuvre d’un système de perception bio-inspiré basé sur l’attention visuelle et auditiveWang, Jingyu 09 January 2015 (has links)
L'objectif principal de cette thèse porte sur la conception d'un système de perception artificiel permettant d'identifier des scènes ou évènements pertinents dans des environnements complexes. Les travaux réalisés ont permis d'étudier et de mettre en œuvre d'un système de perception bio-inspiré basé sur l'attention visuelle et auditive. Les principales contributions de cette thèse concernent la saillance auditive associée à une identification des sons et bruits environnementaux ainsi que la saillance visuelle associée à une reconnaissance d'objets pertinents. La saillance du signal sonore est calculée en fusionnant des informations extraites des représentations temporelles et spectrales du signal acoustique avec une carte de saillance visuelle du spectrogramme du signal concerné. Le système de perception visuelle est quant à lui composé de deux mécanismes distincts. Le premier se base sur des méthodes de saillance visuelle et le deuxième permet d'identifier l'objet en premier plan. D'autre part, l'originalité de notre approche est qu'elle permet d'évaluer la cohérence des observations en fusionnant les informations extraites des signaux auditifs et visuels perçus. Les résultats expérimentaux ont permis de confirmer l'intérêt des méthodes utilisées dans le cadre de l'identification de scènes pertinentes dans un environnement complexe / The main goal of these researches is the design of one artificial perception system allowing to identify events or scenes in a complex environment. The work carried out during this thesis focused on the study and the conception of a bio-inspired perception system based on the both visual and auditory saliency. The main contributions of this thesis are auditory saliency with sound recognition and visual saliency with object recognition. The auditory saliency is computed by merging information from the both temporal and spectral signals with a saliency map of a spectrogram. The visual perception system is based on visual saliency and recognition of foreground object. In addition, the originality of the proposed approach is the possibility to do an evaluation of the coherence between visual and auditory observations using the obtained information from the features extracted from both visual and auditory patters. The experimental results have proven the interest of this method in the framework of scene identification in a complex environment
|
20 |
An online and adaptive signature-based approach for intrusion detection using learning classifier systemsShafi, Kamran, Information Technology & Electrical Engineering, Australian Defence Force Academy, UNSW January 2008 (has links)
This thesis presents the case of dynamically and adaptively learning signatures for network intrusion detection using genetic based machine learning techniques. The two major criticisms of the signature based intrusion detection systems are their i) reliance on domain experts to handcraft intrusion signatures and ii) inability to detect previously unknown attacks or the attacks for which no signatures are available at the time. In this thesis, we present a biologically-inspired computational approach to address these two issues. This is done by adaptively learning maximally general rules, which are referred to as signatures, from network traffic through a supervised learning classifier system, UCS. The rules are learnt dynamically (i.e., using machine intelligence and without the requirement of a domain expert), and adaptively (i.e., as the data arrives without the need to relearn the complete model after presenting each data instance to the current model). Our approach is hybrid in that signatures for both intrusive and normal behaviours are learnt. The rule based profiling of normal behaviour allows for anomaly detection in that the events not matching any of the rules are considered potentially harmful and could be escalated for an action. We study the effect of key UCS parameters and operators on its performance and identify areas of improvement through this analysis. Several new heuristics are proposed that improve the effectiveness of UCS for the prediction of unseen and extremely rare intrusive activities. A signature extraction system is developed that adaptively retrieves signatures as they are discovered by UCS. The signature extraction algorithm is augmented by introducing novel subsumption operators that minimise overlap between signatures. Mechanisms are provided to adapt the main algorithm parameters to deal with online noisy and imbalanced class data. The performance of UCS, its variants and the signature extraction system is measured through standard evaluation metrics on a publicly available intrusion detection dataset provided during the 1999 KDD Cup intrusion detection competition. We show that the extended UCS significantly improves test accuracy and hit rate while significantly reducing the rate of false alarms and cost per example scores than the standard UCS. The results are competitive to the best systems participated in the competition in addition to our systems being online and incremental rule learners. The signature extraction system built on top of the extended UCS retrieves a magnitude smaller rule set than the base UCS learner without any significant performance loss. We extend the evaluation of our systems to real time network traffic which is captured from a university departmental server. A methodology is developed to build fully labelled intrusion detection dataset by mixing real background traffic with attacks simulated in a controlled environment. Tools are developed to pre-process the raw network data into feature vector format suitable for UCS and other related machine learning systems. We show the effectiveness of our feature set in detecting payload based attacks.
|
Page generated in 0.0704 seconds