Coding Electronic Medical Records (EMRs) with diagnosis and procedure codes is an essential task for billing, secondary data analyses, and monitoring health trends. Both speed and accuracy of coding are critical. While coding errors could lead to more patient-side financial burden and misinterpretation of a patient’s well-being, timely coding is also needed to avoid backlogs and additional costs for the healthcare facility. Therefore, it is necessary to develop automated diagnosis and procedure code recommendation methods that can be used by professional medical coders. The main difficulty with developing automated EMR coding methods is the nature of the label space. The standardized vocabularies used for medical coding contain over 10 thousand codes. The label space is large, and the label distribution is extremely unbalanced - most codes occur very infrequently, with a few codes occurring several orders of magnitude more than others. A few codes never occur in training dataset at all. In this work, we present three methods to handle the large unbalanced label space. First, we study how to augment EMR training data with biomedical data (research articles indexed on PubMed) to improve the performance of standard neural networks for text classification. PubMed indexes more than 23 million citations. Many of the indexed articles contain relevant information about diagnosis and procedure codes. Therefore, we present a novel method of incorporating this unstructured data in PubMed using transfer learning. Second, we combine ideas from metric learning with recent advances in neural networks to form a novel neural architecture that better handles infrequent codes. And third, we present new methods to predict codes that have never appeared in the training dataset. Overall, our contributions constitute advances in neural multi-label text classification with potential consequences for improving EMR coding.

Expert labeling, tagging, and assessment are far more costly than the processes of collecting raw data. Generative modeling is a very powerful tool to tackle this real-world problem. It is shown here how these models can be used to allow for semi-supervised learning that performs very well in label-deficient conditions. The foundation for the work in this dissertation is built upon visualizing generative models' latent spaces to gain deeper understanding of data, analyze faults, and propose solutions. A number of novel ideas and approaches are presented to improve single-label classification. This dissertation's main focus is on extending semi-supervised Deep Generative Models for solving the multi-label problem by proposing unique mathematical and programming concepts and organization. In all naive mixtures, using multiple labels is detrimental and causes each label's predictions to be worse than models that utilize only a single label. Examining latent spaces reveals that in many cases, large regions in the models generate meaningless results. Enforcing a priori independence is essential, and only when applied can multi-label models outperform the best single-label models. Finally, a novel learning technique called open-book learning is described that is capable of surpassing the state-of-the-art classification performance of generative models for multi-labeled, semi-supervised data sets.

This thesis proposes a series of multi-label learning algorithms for classification and feature selection implemented on the Apache Spark distributed computing model. Five approaches for determining the optimal architecture to speed up multi-label learning methods are presented. These approaches range from local parallelization using threads to distributed computing using independent or shared memory spaces. It is shown that the optimal approach performs hundreds of times faster than the baseline method. Three distributed multi-label k nearest neighbors methods built on top of the Spark architecture are proposed: an exact iterative method that computes pair-wise distances, an approximate tree-based method that indexes the instances across multiple nodes, and an approximate local sensitive hashing method that builds multiple hash tables to index the data. The results indicated that the predictions of the tree-based method are on par with those of an exact method while reducing the execution times in all the scenarios. The aforementioned method is then used to evaluate the quality of a selected feature subset. The optimal adaptation for a multi-label feature selection criterion is discussed and two distributed feature selection methods for multi-label problems are proposed: a method that selects the feature subset that maximizes the Euclidean norm of individual information measures, and a method that selects the subset of features maximizing the geometric mean. The results indicate that each method excels in different scenarios depending on type of features and the number of labels. Rigorous experimental studies and statistical analyses over many multi-label metrics and datasets confirm that the proposals achieve better performances and provide better scalability to bigger data than the methods compared in the state of the art.

MEDLINE, a large and constantly increasing collection of biomedical article references, has been the source of numerous investigations related to textual information retrieval and knowledge capture, including article categorization, bibliometric analysis, semantic query answering, and biological concept recognition and relationship extraction. This dissertation discusses the design and development of novel methods that contribute to the tasks of document categorization and relationship extraction. The two investigations result in a fast tool for building descriptive models capable of categorizing documents to multiple labels and a highly effective method able to extract broad range of relationships between entities embedded in text. Additionally, an application that aims at representing the extracted knowledge in a strictly defined but highly expressive structure of ontology is presented. The classification of documents is based on an idea of building association rules that consist of frequent patterns of words appearing in documents and classes these patterns are likely to be assigned to. The process of building the models is based on a tree enumeration technique and dataset projection. The resulting algorithm offers two different tree traversing strategies, breadth-first and depth-first. The classification scenario involves the use of two alternative thresholding strategies based on either the document-independent confidence of the rules or a similarity measure between a rule and a document. The presented classification tool is shown to perform faster than other methods and is the first associative-classification solution to incorporate multiple classes and the information about recurrence of words in documents. The extraction of relations between entities embedded in text involves the utilization of the output of a constituent parser and a set of manually developed tree-like patterns. Both serve as the input of a novel algorithm that solves the newly formulated problem of constrained constituent tree inclusion with regular expression matching. The proposed relation extraction method is demonstrated to be parser-independent and outperforms in terms of effectiveness dependency-parser-based and machine-learning-based solutions. The extracted knowledge is further embedded in an existing ontology, which together with the structure-driven modification of the ontology results in a comprehensible, inference-consistent knowledge base constituting a tangible representation of knowledge and a potential component of applications such as semantically enhanced query answering systems.

No description available.

In many real-world applications of supervised learning, only a limited number of labeled examples are available because the cost of obtaining high-quality examples is high. Even with a relatively large number of labeled examples, the learning problem may still suffer from limited supervision as the complexity of the prediction function increases. Therefore, learning with limited supervision presents a major challenge to machine learning. With the goal of supervision reduction, this thesis studies the representation, discovery and incorporation of extra input and output information in learning. Information about the input space can be encoded by regularization. We first design a semi-supervised learning method for text classification that encodes the correlation of words inferred from seemingly irrelevant unlabeled text. We then propose a multi-task learning framework with a matrix-normal penalty, which compactly encodes the covariance structure of the joint input space of multiple tasks. To capture structure information that is more general than covariance and correlation, we study a class of regularization penalties on model compressibility. Then we design the projection penalty, which encodes the structure information from a dimension reduction while controlling the risk of information loss. Information about the output space can be exploited by error correcting output codes. Using the composite likelihood view, we propose an improved pairwise coding for multi-label classification, which encodes pairwise label density (as opposed to label comparisons) and decodes using variational methods. We then investigate problemdependent codes, where the encoding is learned from data instead of being predefined. We first propose a multi-label output code using canonical correlation analysis, where predictability of the code is optimized. We then argue that both discriminability and predictability are critical for output coding, and propose a max-margin formulation that promotes both discriminative and predictable codes. We empirically study our methods in a wide spectrum of applications, including document categorization, landmine detection, face recognition, brain signal classification, handwritten digit recognition, house price forecasting, music emotion prediction, medical decision, email analysis, gene function classification, outdoor scene recognition, and so forth. In all these applications, our proposed methods for encoding input and output information lead to significantly improved prediction performance.

A tarefa de classificação em Aprendizado de Máquina consiste da criação de modelos computacionais capazes de identificar automaticamente a classe de objetos pertencentes a um domínio pré-definido a partir de um conjunto de exemplos cuja classe é conhecida. Existem alguns cenários de classificação nos quais cada objeto pode estar associado não somente a uma classe, mas a várias classes ao mesmo tempo. Adicionalmente, nesses cenários denominados multirrótulo, as classes podem ser organizadas em uma taxonomia que representa as relações de generalização e especialização entre as diferentes classes, definindo uma hierarquia de classes, o que torna a tarefa de classificação ainda mais específica, denominada classificação hierárquica. Os métodos utilizados para a construção desses modelos de classificação são complexos e dependem fortemente da disponibilidade de uma quantidade expressiva de exemplos previamente classificados. Entretanto, para muitas aplicações é difícil encontrar um número significativo desses exemplos. Além disso, com poucos exemplos, os algoritmos de aprendizado supervisionado não são capazes de construir modelos de classificação eficazes. Nesses casos, é possível utilizar métodos de aprendizado semissupervisionado, cujo objetivo é aprender as classes do domínio utilizando poucos exemplos conhecidos conjuntamente com um número considerável de exemplos sem a classe especificada. Neste trabalho são propostos, entre outros, métodos que fazem uso do aprendizado semissupervisionado baseado em desacordo coperspectiva, tanto para a tarefa de classificação multirrótulo plana quanto para a tarefa de classificação hierárquica. São propostos, também, outros métodos que utilizam o aprendizado ativo com intuito de melhorar a performance de algoritmos de classificação semissupervisionada. Além disso, são propostos dois métodos para avaliação de algoritmos multirrótulo e hierárquico, os quais definem estratégias para identificação dos multirrótulos majoritários, que são utilizados para calcular os valores baseline das medidas de avaliação. Foi desenvolvido um framework para realizar a avaliação experimental da classificação hierárquica, no qual foram implementados os métodos propostos e um módulo completo para realizar a avaliação experimental de algoritmos hierárquicos. Os métodos propostos foram avaliados e comparados empiricamente, considerando conjuntos de dados de diversos domínios. A partir da análise dos resultados observa-se que os métodos baseados em desacordo não são eficazes para tarefas de classificação complexas como multirrótulo e hierárquica. Também é observado que o problema central de degradação do modelo dos algoritmos semissupervisionados agrava-se nos casos de classificação multirrótulo e hierárquica, pois, nesses casos, há um incremento nos fatores responsáveis pela degradação nos modelos construídos utilizando aprendizado semissupervisionado baseado em desacordo coperspectiva / In machine learning, the task of classification consists on creating computational models that are able to automatically identify the class of objects belonging to a predefined domain from a set of examples whose class is known a priori. There are some classification scenarios in which each object can be associated to more than one class at the same time. Moreover, in such multilabeled scenarios, classes can be organized in a taxonomy that represents the generalization and specialization relationships among the different classes, which defines a class hierarchy, making the classification task, known as hierarchical classification, even more specific. The methods used to build such classification models are complex and highly dependent on the availability of an expressive quantity of previously classified examples. However, for a large number of applications, it is difficult to find a significant number of such examples. Moreover, when few examples are available, supervised learning algorithms are not able to build efficient classification models. In such situations it is possible to use semi-supervised learning, whose aim is to learn the classes of the domain using a few classified examples in conjunction to a considerable number of examples with no specified class. In this work, we propose methods that use the co-perspective disagreement based learning approach for both, the flat multilabel classification and the hierarchical classification tasks, among others. We also propose other methods that use active learning, aiming at improving the performance of semi-supervised learning algorithms. Additionally, two methods for the evaluation of multilabel and hierarchical learning algorithms are proposed. These methods define strategies for the identification of the majority multilabels, which are used to estimate the baseline evaluation measures. A framework for the experimental evaluation of the hierarchical classification was developed. This framework includes the implementations of the proposed methods as well as a complete module for the experimental evaluation of the hierarchical algorithms. The proposed methods were empirically evaluated considering datasets from various domains. From the analysis of the results, it can be observed that the methods based on co-perspective disagreement are not effective for complex classification tasks, such as the multilabel and hierarchical classification. It can also be observed that the main degradation problem of the models of the semi-supervised algorithms worsens for the multilabel and hierarchical classification due to the fact that, for these cases, there is an increase in the causes of the degradation of the models built using semi-supervised learning based on co-perspective disagreement

Made available in DSpace on 2016-12-23T14:33:42Z (GMT). No. of bitstreams: 1 Dissertacao de Caribe Zampirolli de Souza.pdf: 1221547 bytes, checksum: 1e22f89c93c3423e4143b9ac13eeb1c6 (MD5) Previous issue date: 2010-08-27 / A multi-label text categorization system typically computes degrees of belief when it comes to the categories of a pre-defined set, orders the categories by degree of belief, and attributes to the document categories with a higher degree of belief to determined threshold cut. It would be ideal if the degree of belief could inform the probability of the document be part of this category. Unfortunately, there isn t a categorization system that computes such probabilities and to map degrees of belief in probabilities is still a problem that isn`t well explored in IR. In this paper we propose a method based on Bayes rules to map degrees of belief in terms of multi-label text measures of categorization. There are other contributions in this work such as an strategy to determine the limits of threshold cut based on bayesian cut (BCut) and a variant for PBCut (position based bayesian CUT ). As an experience, we evaluated the impact of the proposed methods when performing the two techniques of the multi-label text categorization. The first technique is called knearest neighbor multi-label (ML-KNN) and the second technique is called VG-RAM weightless Neural Networks. Theses evaluations were made in the context of the categorization of economic activities description of Brazilian enterprises, according to the Economic Activities Classification in Brazil (CNAE). In this work we also investigated the impact in the performance of multi-label text categorization of the three cut methods commonly used in the IR literature: RCut, PCut, SCut and RTCut. Moreover, we propose a new variant for the so called PCut* and a new variant for SCut*. Finally, this work shows that the cut approach proposed, BCut and PBCut, produces a categorization performance superior to the other strategies presented in the literature of IR / Dado um documento de entrada, um sistema de categorização multi-rótulo de texto tipicamente computa graus de crença para as categorias de um conjunto pré-definido, ordena as categorias por grau de crença, e atribui ao documento as categorias com grau de crença superior a um determinado limiar de poda. Idealmente, o grau de crença deveria informar a probabilidade do documento de fato pertencer à categoria. Infelizmente, ainda não existem categorizadores que computam tais probabilidades e mapear graus de crença em probabilidades é um problema ainda pouco explorado na área de RI. Neste trabalho, propomos um método baseado na regra de Bayes para mapear graus de crença em medidas de certeza de categorização multi-rótulo de texto. Propomos também uma estratégia para determinar limiares de poda baseada na medida de certeza de categorização - bayesian cut (BCut) - e uma variante para BCut - position based bayesian CUT (PBCut). Avaliamos experimentalmente o impacto dos métodos propostos no desempenho de duas técnicas de categorização multi-rótulo de texto, k-vizinhos mais próximos multi-rótulo (MLkNN) e rede neural sem peso do tipo VG-RAM com correlação de dados (VG-RAM WNNCOR), no contexto da categorização de descrições de atividades econômicas de empresas brasileiras segundo a Classificação Nacional de Atividades Econômicas (CNAE). Investigamos também o impacto no desempenho de categorização multi-rótulo de texto de três métodos de poda comumente usados na literatura de RI - RCut, PCut, e SCut e uma variante de RCut - RTCut. Além disso, propomos novas variantes para PCut e SCut PCut* e SCut*, respectivamente para tratar problemas existentes nestas abordagens. Nossos resultados experimentais mostram que, usando nosso método de geração de medidas de certeza de categorização, é possível prever o quão certo está o categorizador de que as ategorias por ele preditas são de fato pertinentes para um dado documento. Nossos resultados mostram também que o uso de nossas estratégias de poda BCut e PBCut produz desempenho de categorização superior ao de todas as outras estratégias consideradas em termos de precisão

Métodos tradicionais de aprendizado supervisionado, chamados de aprendizado monorrótulo, consideram que cada exemplo do conjunto de dados rotulados está associado a um único rótulo. No entanto, existe uma crescente quantidade de aplicações que lidam com exemplos que estão associados a múltiplos rótulos. Essas aplicações requerem métodos de aprendizado multirrótulo. Esse cenário de aprendizado introduz novos desafios que demandam abordagens diferentes daquelas tradicionalmente utilizadas no aprendizado monorrótulo. O custo associado ao processo de rotulação de exemplos, um problema presente em aprendizado monorrótulo, é ainda mais acentuado no contexto multirrótulo. O desenvolvimento de métodos para reduzir esse custo representa um desafio de pesquisa nessa área. Além disso, novos métodos de aprendizado também devem ser desenvolvidos para, entre outros objetivos, considerar a dependência de rótulos: uma nova característica presente no aprendizado multirrótulo. Há um consenso na comunidade de que métodos de aprendizado multirrótulo têm a capacidade de usufruir de melhor eficácia preditiva quando considerada a dependência de rótulos. Os principais objetivos deste trabalho estão relacionados a esses desafios: reduzir o custo do processo de rotulação de exemplos; e desenvolver métodos de aprendizado que explorem a dependência de rótulos. No primeiro caso, entre outras contribuições, um novo método de aprendizado ativo, chamado score dev, é proposto para reduzir os custos associados ao processo de rotulação multirrótulo. Resultados experimentais indicam que o método score dev é superior a outros métodos em vários domínios. No segundo caso, um método para identificar dependência de rótulos, chamado UBC, é proposto, bem como o BR+, um método para explorar essa característica. O método BR+ apresenta resultados superiores a métodos considerados estado da arte / Traditional supervised learning methods, called single-label learning, consider that each example from a labeled dataset is associated with only one label. However, an increasing number of applications deals with examples that are associated with multiple labels. These applications require multi-label learning methods. This learning scenario introduces new challenges and demands approaches that are different from those traditionally used in single-label learning. The cost of labeling examples, a problem in single-label learning, is even higher in the multi-label context. Developing methods to reduce this cost represents a research challenge in this area. Moreover, new learning methods should also be developed to, among other things, consider the label dependency: a new characteristic present in multi-label learning problems. Furthermore, there is a consensus in the community that multi-label learning methods are able to improve their predictive performance when label dependency is considered. The main aims of this work are related to these challenges: reducing the cost of the labeling process; and developing multi-label learning methods to explore label dependency. In the first case, as well as other contributions, a new multi-label active learning method, called score dev, is proposed to reduce the multi-labeling processing costs. Experimental results show that score dev outperforms other methods in many domains. In the second case, a method to identify label dependency, called UBC, is proposed, as well as BR+, a method to explore this characteristic. Results show that the BR+ method outperforms other state-of-the-art methods

Muitos problemas de classificação descritos na literatura de aprendizado de máquina e mineração de dados dizem respeito à classificação em que cada exemplo pertence a um único rótulo. Porém, vários problemas de classificação, principalmente no campo de Bioinformática são associados a mais de um rótulo; esses problemas são conhecidos como problemas de classificação multirrótulo. O princípio básico da classificação multirrótulo é similar ao da classificação tradicional (que possui um único rótulo), sendo diferenciada no número de rótulos a serem preditos, na qual há dois ou mais rótulos. Na área da Bioinformática muitos problemas são compostos por uma grande quantidade de rótulos em que cada exemplo pode estar associado. Porém, algoritmos de classificação tradicionais são incapazes de lidar com um conjunto de exemplos mutirrótulo, uma vez que esses algoritmos foram projetados para predizer um único rótulo. Uma solução mais simples é utilizar o método conhecido como método Binary Relevance. Porém, estudos mostraram que tal abordagem não constitui uma boa solução para o problema da classificação multirrótulo, pois cada classe é tratada individualmente, ignorando as possíveis relações entre elas. Dessa maneira, o objetivo dessa pesquisa foi propor uma nova adaptação do método Binary Relevance que leva em consideração relações entre os rótulos para tentar minimizar sua desvantagem, além de também considerar a capacidade de interpretabilidade do modelo gerado, não só o desempenho. Os resultados experimentais mostraram que esse novo método é capaz de gerar árvores que relacionam os rótulos correlacionados e também possui um desempenho comparável ao de outros métodos, obtendo bons resultados usando a medida-F. / Many classification problems described in the literature on Machine Learning and Data Mining relate to the classification in which each example belongs to a single class. However, many classification problems, especially in the field of Bioinformatics, are associated with more than one class; these problems are known as multi-label classification problems. The basic principle of multi-label classification is similar to the traditional classification (single label), and distinguished by the number of classes to be predicted, in this case, in which there are two or more labels. In Bioinformatics many problems are composed of a large number of labels that can be associated with each example. However, traditional classification algorithms are unable to cope with a set of multi-label examples, since these algorithms are designed to predict a single label. A simpler solution is to use the method known as Binary Relevance. However, studies have shown that this approach is not a good solution to the problem of multi-label classification because each class is treated individually, ignoring possible relations between them. Thus, the objective of this research was to propose a new adaptation of Binary Relevance method that took into account relations between labels trying to minimize its disadvantage, and also consider the ability of interpretability of the model generated, not just its performance. The experimental results show that this new method is capable of generating trees that relate labels and also has a performance comparable to other methods, obtaining good results using F-measure.