Rule Extraction From Support Vector Machines

Support vector machines (SVMs) are one of the most active research areas in machine learning. SVMs have shown good performance in a number of applications, including text and image classification. However, the learning capability of SVMs comes at a cost – an inherent inability to explain in a comprehensible form, the process by which a learning result was reached. Hence, the situation is similar to neural networks, where the apparent lack of an explanation capability has led to various approaches aiming at extracting symbolic rules from neural networks. For SVMs to gain a wider degree of acceptance in fields such as medical diagnosis and security sensitive areas, it is desirable to offer an explanation capability. User explanation is often a legal requirement, because it is necessary to explain how a decision was reached or why it was made. This book provides an overview of the field and introduces a number of different approaches to extracting rules from support vector machines developed by key researchers. In addition, successful applications are outlined and future research opportunities are discussed. The book is an important reference for researchers and graduate students, and since it provides an introduction to the topic, it will be important in the classroom as well. Because of the significance of both SVMs and user explanation, the book is of relevance to data mining practitioners and data analysts.

Doctoral Thesis / Dissertation from the year 2010 in the subject Computer Science - Applied, grade: none, course: Department of Computers and Information Sciences - Ph.D., language: English, abstract: Although Support Vector Machines have been used to develop highly accurate classification and regression models in various real-world problem domains, the most significant barrier is that SVM generates black box model that is difficult to understand. The procedure to convert these opaque models into transparent models is called rule extraction. This thesis investigates the task of extracting comprehensible models from trained SVMs, thereby alleviating this limitation. The primary contribution of the thesis is the proposal of various algorithms to overcome the significant limitations of SVM by taking a novel approach to the task of extracting comprehensible models. The basic contribution of the thesis are systematic review of literature on rule extraction from SVM, identifying gaps in the literature and proposing novel approaches for addressing the gaps. The contributions are grouped under three classes, decompositional, pedagogical and eclectic/hybrid approaches. Decompositional approach is closely intertwined with the internal workings of the SVM. Pedagogical approach uses SVM as an oracle to re-label training examples as well as artificially generated examples. In the eclectic/hybrid approach, a combination of these two methods is adopted. The thesis addresses various problems from the finance domain such as bankruptcy prediction in banks/firms, churn prediction in analytical CRM and Insurance fraud detection. Apart from this various benchmark datasets such as iris, wine and WBC for classification problems and auto MPG, body fat, Boston housing, forest fires and pollution for regression problems are also tested using the proposed appraoch. In addition, rule extraction from unbalanced datasets as well as from active learning based approaches has been explored. For

Doctoral Thesis / Dissertation from the year 2010 in the subject Computer Science - Applied, grade: none, , course: Department of Computers and Information Sciences - Ph.D., language: English, abstract: Although Support Vector Machines have been used to develop highly accurate classification and regression models in various real-world problem domains, the most significant barrier is that SVM generates black box model that is difficult to understand. The procedure to convert these opaque models into transparent models is called rule extraction. This thesis investigates the task of extracting comprehensible models from trained SVMs, thereby alleviating this limitation. The primary contribution of the thesis is the proposal of various algorithms to overcome the significant limitations of SVM by taking a novel approach to the task of extracting comprehensible models. The basic contribution of the thesis are systematic review of literature on rule extraction from SVM, identifying gaps in the literature and proposing novel approaches for addressing the gaps. The contributions are grouped under three classes, decompositional, pedagogical and eclectic/hybrid approaches. Decompositional approach is closely intertwined with the internal workings of the SVM. Pedagogical approach uses SVM as an oracle to re-label training examples as well as artificially generated examples. In the eclectic/hybrid approach, a combination of these two methods is adopted. The thesis addresses various problems from the finance domain such as bankruptcy prediction in banks/firms, churn prediction in analytical CRM and Insurance fraud detection. Apart from this various benchmark datasets such as iris, wine and WBC for classification problems and auto MPG, body fat, Boston housing, forest fires and pollution for regression problems are also tested using the proposed appraoch. In addition, rule extraction from unbalanced datasets as well as from active learning based approaches has been explored. For classification problems, various rule extraction methods such as FRBS, DT, ANFIS, CART and NBTree have been utilized. Additionally for regression problems, rule extraction methods such as ANFIS, DENFIS and CART have also been employed. Results are analyzed using accuracy, sensitivity, specificity, fidelity, AUC and t-test measures. Proposed approaches demonstrate their viability in extracting accurate, effective and comprehensible rule sets in various benchmark and real world problem domains across classification and regression problems. Future directions have been indicated to extend the approaches to newer variations of SVM as well as to other problem domains.

Based on ideas from Support Vector Machines (SVMs), Learning To Classify Text Using Support Vector Machines presents a new approach to generating text classifiers from examples. The approach combines high performance and efficiency with theoretical understanding and improved robustness. In particular, it is highly effective without greedy heuristic components. The SVM approach is computationally efficient in training and classification, and it comes with a learning theory that can guide real-world applications. Learning To Classify Text Using Support Vector Machines gives a complete and detailed description of the SVM approach to learning text classifiers, including training algorithms, transductive text classification, efficient performance estimation, and a statistical learning model of text classification. In addition, it includes an overview of the field of text classification, making it self-contained even for newcomers to the field. This book gives a concise introduction to SVMs for pattern recognition, and it includes a detailed description of how to formulate text-classification tasks for machine learning.

When discussing classification, support vector machines are known to be a capable and efficient technique to learn and predict with high accuracy within a quick time frame. Yet, their black box means to do so make the practical users quite circumspect about relying on it, without much understanding of the how and why of its predictions. The question raised in this book is how can this ‘masked hero’ be made more comprehensible and friendly to the public: provide a surrogate model for its hidden optimization engine, replace the method completely or appoint a more friendly approach to tag along and offer the much desired explanations? Evolutionary algorithms can do all these and this book presents such possibilities of achieving high accuracy, comprehensibility, reasonable runtime as well as unconstrained performance.

Computational optimization is an important paradigm with a wide range of applications. In virtually all branches of engineering and industry, we almost always try to optimize something - whether to minimize the cost and energy consumption, or to maximize profits, outputs, performance and efficiency. In many cases, this search for optimality is challenging, either because of the high computational cost of evaluating objectives and constraints, or because of the nonlinearity, multimodality, discontinuity and uncertainty of the problem functions in the real-world systems. Another complication is that most problems are often NP-hard, that is, the solution time for finding the optimum increases exponentially with the problem size. The development of efficient algorithms and specialized techniques that address these difficulties is of primary importance for contemporary engineering, science and industry. This book consists of 12 self-contained chapters, contributed from worldwide experts who are working in these exciting areas. The book strives to review and discuss the latest developments concerning optimization and modelling with a focus on methods and algorithms for computational optimization. It also covers well-chosen, real-world applications in science, engineering and industry. Main topics include derivative-free optimization, multi-objective evolutionary algorithms, surrogate-based methods, maximum simulated likelihood estimation, support vector machines, and metaheuristic algorithms. Application case studies include aerodynamic shape optimization, microwave engineering, black-box optimization, classification, economics, inventory optimization and structural optimization. This graduate level book can serve as an excellent reference for lecturers, researchers and students in computational science, engineering and industry.

This book constitutes the refereed proceedings of the 12th International Conference on Rough Sets, Fuzzy Sets, Data Mining, and Granular Computing, RSFDGrC 2009, held in Delhi, India in December 2009 in conjunction with the Third International Conference on Pattern Recognition and Machine Intelligence, PReMI 2009. RSFDGrC 2009 is the core component of a broader Rough Set Year in India initiative, RSIndia09. The 56 revised full papers presented together with 6 invited papers and a report on the Rough Set Year in India 2009 project were carefully reviewed and selected from a total of 130 submissions. The papers are organized in topical sections on foundations of rough sets and beyond; rought set algorithms and applications; fuzzy set foundations and applications; data mining and knowledge discovery; clustering and current trends in computing; and information retrieval and text mining.

This book provides a unified approach for developing a fuzzy classifier and explains the advantages and disadvantages of different classifiers through extensive performance evaluation of real data sets. It thus offers new learning paradigms for analyzing neural networks and fuzzy systems, while training fuzzy classifiers. Function approximation is also treated and function approximators are compared.