Computational intelligence for semantic knowledge management, 1st ed , giovanni acampora, witold pedrycz, athanasios v vasilakos, autilia vitiello, 2020 313

150 1 0
  • Loading ...
1/150 trang
Tải xuống

Thông tin tài liệu

Ngày đăng: 08/05/2020, 06:57

Studies in Computational Intelligence 837 Giovanni Acampora Witold Pedrycz Athanasios V. Vasilakos Autilia Vitiello   Editors Computational Intelligence for Semantic Knowledge Management New Perspectives for Designing and Organizing Information Systems Studies in Computational Intelligence Volume 837 Series Editor Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland The series “Studies in Computational Intelligence” (SCI) publishes new developments and advances in the various areas of computational intelligence—quickly and with a high quality The intent is to cover the theory, applications, and design methods of computational intelligence, as embedded in the fields of engineering, computer science, physics and life sciences, as well as the methodologies behind them The series contains monographs, lecture notes and edited volumes in computational intelligence spanning the areas of neural networks, connectionist systems, genetic algorithms, evolutionary computation, artificial intelligence, cellular automata, self-organizing systems, soft computing, fuzzy systems, and hybrid intelligent systems Of particular value to both the contributors and the readership are the short publication timeframe and the world-wide distribution, which enable both wide and rapid dissemination of research output The books of this series are submitted to indexing to Web of Science, EI-Compendex, DBLP, SCOPUS, Google Scholar and Springerlink More information about this series at http://www.springer.com/series/7092 Giovanni Acampora Witold Pedrycz Athanasios V Vasilakos Autilia Vitiello • • • Editors Computational Intelligence for Semantic Knowledge Management New Perspectives for Designing and Organizing Information Systems 123 Editors Giovanni Acampora Department of Physics “Ettore Pancini” University of Naples Federico II Naples, Italy Athanasios V Vasilakos Department of Computer Science, Electrical and Space Engineering Luleå University of Technology Skellefteå, Sweden Witold Pedrycz Department of Electrical and Computer Engineering University of Alberta Edmonton, AB, Canada Autilia Vitiello Department of Physics “Ettore Pancini” University of Naples Federico II Naples, Italy ISSN 1860-949X ISSN 1860-9503 (electronic) Studies in Computational Intelligence ISBN 978-3-030-23758-5 ISBN 978-3-030-23760-8 (eBook) https://doi.org/10.1007/978-3-030-23760-8 © Springer Nature Switzerland AG 2020 This work is subject to copyright All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed The use of general descriptive names, registered names, trademarks, service marks, etc in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland If at first you don’t succeed, try, try, try again —William Edward Hickson Preface In last years, knowledge management is becoming a big challenge especially due to the massive quantity of data available in digital form on the web or in large enterprises Hence, there is a large interest in research and industrial activities devoted to design and develop advanced Knowledge Management Systems (KMSs) KMSs are systematic frameworks destined to capture, acquire, organize, and communicate both tacit and explicit knowledge in the most effective way More recently, the Semantic Web perspective has added to KMSs a new capability, that is, to offer more intelligent services by facilitating machine understanding of content Hence, there is the birth of the so-called Semantic Knowledge Management (SKM) including a collection of methods, paradigms, and technologies for efficiently supporting the representation and management of tangible knowledge assets Thanks to its flexibility, scalability, and robustness in the representation of information, SKM is enabling the design and development of innovative information systems in distributed environments, ensuring: • Semantic interoperability and integration of data, information, and processes, through an ontological representation of information and • Efficient extraction of interesting information from data in large databases by means of sophisticated Semantic Knowledge Discovery techniques However, in spite of their numerous benefits, SKM methods are not yet able to address some of the problems that intrinsically characterize the representation of knowledge, such as the vagueness and uncertainty of information Computational Intelligence (CI) methodologies, due to their natural inclination to deal with imprecision and partial truth, are opening new positive scenarios for designing innovative SKM architectures For instance, fuzzy logic has inspired the development of different fuzzy extensions of several XML-based languages to enhance the description power of current languages for Semantic Web At the same way, biological-inspired optimization methods such as genetic algorithms and particle swarm optimization have been witnessed in reducing the complexity of several computational problems inherent the semantic representation of information, such vii viii Preface as ontology alignment and matching, query processing, semantic storage, and web-scale reasoning This edited book volume is primarily intended to be a collection of chapters written by experts in the field of the usage of CI methods to the context of the SKM The book is organized into six chapters A summary of the chapters follow: • Fuzzy logic is applied in the chapter by Raciel Yera, Jorge Castro, and Luis Martínez entitled “Natural Noise Management in Recommender Systems using Fuzzy Tools” for dealing with identification of natural noise in Recommender Systems (RSs) RSs are information filtering systems aimed to predict the probability of a user preferring a particular item out of a given set of items These systems require the elicitation of user preferences, which are not always precise because there are external factors such as human errors or the inherent vagueness associated with human beings Such imprecise behaviors are identified as Natural Noise (NN), and can negatively affect the RS performance The authors propose two fuzzy models for NN management in a flexible way, which guarantee robust modeling of the uncertainty associated with the user profiles Two case studies are developed to show that the proposed approaches lead to improvements in the accuracy of RSs • Statistical- and semantic-based approaches are used in the chapter by Loredana Caruccio, Vincenzo Deufemia, Salvatore Esposito, and Giuseppe Polese entitled “Combining Collaborative Filtering and Semantic-based Techniques to Recommend Components for Mashup Design” in order to support web mashup development Mashups merge data from different web sources to create new functionalities, and hence it requires to manage a large amount of heterogeneous knowledge Currently, researchers are investigating both semantic and statistical approaches to detect mashup components that best match user needs In this chapter, the authors present a hybrid recommendation approach combining both the statistical nature of collaborative filtering and semantic methods to select the mashups on the web that are more suitable for satisfying user needs A prototype of the proposed approach is used to prove its validity during three experimental sessions • Semantic maps are used in the chapter by Francesco Camastra, Angelo Ciaramella, Antonio Maratea, Le Hoang Son, and Antonino Staiano entitled “Semantic Maps for Knowledge Management of Web and Social Information” in order to extract potentially useful knowledge from World Wide Web (WWW) Due to the continuous increase in volume and to the mainly unstructured nature of most of the data stored in the WWW, several challenging problems have emerged, the most important being how to find relevant information for a specific task In the chapter, the authors address two representative tasks: first, to provide a compact and structured representation of the main concepts in a Web document; second, to represent and synthesize the information content of Twitter conversations in the form of semantic maps The results of the experiments involving the corpus Reuters and real data show good performance of the semantic proposed approaches Preface ix • Local search meta-heuristics are applied in the chapter by Giovanni Acampora and Autilia Vitiello entitled “A Study on Local Search Meta-heuristics for Ontology Alignment” to reconcile different knowledge sources Currently, the most popular representation methods for the knowledge are the ontologies, however, the variety of ways that a domain can be conceptualized results in the development of heterogeneous ontologies with overlapping parts In order to address this problem, a so-called ontology alignment process is required In the chapter, the authors propose to implement an ontology alignment process, for the first time, by means of local search algorithms As shown by the results of a set of experiments involving well-known benchmarks, Tabu search is the best performer among the compared local search meta-heuristics • Decision trees are used in the chapter by Sriparna Saha, Shreyasi Datta, and Amit Konar entitled “Decision Tree Based Single Person Gesture Recognition” in order to classify emotions by managing knowledge captured from human behaviors Classifying emotions starting from human gestures can be utilized to control a machine according to the human emotional state Human gestures for the present work are captured using a Kinect Sensor which tracks the skeleton of the person standing in front of it within a finite amount of distance using a set of visible and IR cameras The results of the experiments conducted on 10 subjects show the good performance of the proposed approach • The integration between type-2 fuzzy logic and evolutionary computation is used in the chapter by Sriparna Saha, Pratyusha Rakshit, and Amit Konar entitled “Modified Type-2 Fuzzy Gesture Space Induced Physical Disorder Recognition” in order to manage information captured by human gestures and recognize physical disorders The gestural features of a subject suffering from the same physical disorder exhibit wide deviations for different instances This fluctuation is the main source of uncertainty in the physical disorder recognition problem The authors address this problem by means of Type-2 Fuzzy Sets Moreover, Type-2 fuzzy sets are formulated by solving an optimization problem by means of the Artificial Bee Colony As shown the results of the experimental session, the proposed approach improves the state of the art Before concluding, we wish to thank various people for their contribution to this book First, we would like to express our sincere thanks to the authors of the chapters for having made available their experiences related to their research and also for carefully addressing reviewers’ comments In addition, we are indebted to the reviewers for providing useful comments on the chapters Besides, our thanks are due to Springer for publishing this book and for assisting us during the different steps of the publication process Lastly, we are grateful to our families for their continuous support x Preface We strongly hope this book will stimulate and support the activities of researchers in the field of the computational intelligence and in the semantic web area Happy reading! Naples, Italy Edmonton, Canada Skellefteå, Sweden Naples, Italy October 2018 Giovanni Acampora Witold Pedrycz Athanasios V Vasilakos Autilia Vitiello Modified Type-2 Fuzzy Gesture Space Induced Physical Disorder Recognition Table (continued) # Subject Primary membership 36 0.8568 37 0.7810 38 0.9063 39 0.8494 40 0.9332 41 0.8771 42 0.8332 43 0.9426 44 0.9150 45 0.8291 46 0.9090 47 0.9345 48 0.8029 49 0.7853 50 0.7941 Modified lower membership Modified lower membership Modified FOU Lower membership Upper membership 121 Secondary membership Modified Membership 0.8547 0.9222 0.7871 0.8194 0.7813 0.7962 0.7972 0.8960 0.9020 0.8934 0.7914 0.9188 0.8336 0.8492 0.8917 0.9034 0.9013 0.9001 0.8971 0.9028 0.8975 0.8916 0.9227 0.9194 0.9048 0.9011 0.9252 0.8925 0.8921 0.8992 0.8871 0.9268 [0.8871, 0.9268] physical disorder i.e., for c = [1, 16] Thus for 15 features, altogether 240 such tables are obtained Table provides the results of individual range in primary membership for each feature experimented under different gestural conditions For example, the entry (0.88 − 0.93) corresponding to the row Osteoarthritis hand pain at the left joint while sitting and column DERL , gives an idea about the extent of the DERL for the unknown subject matches with known subjects from the physical disorder class Osteoarthritis hand pain at the left joint while sitting The results of computing fuzzy meet operation over the range of individual features taken from gestural expressions of the subjects under the same gestural condition are given in Table The average of the ranges along with its centre value is also given in Table It is observed that the centre has the largest value (= 0.47) for the physical disorder: Frozen shoulder at the right joint while standing Right Standing Right Standing Right Left Right Sitting Standing Standing Left Left Standing Plantar fasci- Sitting itis Right Sitting Left Left Standing Frozen shoul- Sitting der Right Left Sitting Osteoarthritis Sitting hand [0.01, 0.73] [0.02, 0.76] [0.00, 0.07] [0.00, 0.07] [0.78, 0.96] [0.78, 0.98] [0.01, 0.77] [0.01, 0.81] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.11] [0.00, 0.01] [0.05, 0.44] [0.00, 0.38] [0.02, 0.67] [0.01, 0.61] [0.05, 0.50] [0.06, 0.46] [0.00, 0.45] [0.00, 0.54] [0.88, 0.93] [0.88, 0.93] DERL [0.00, 0.27] [0.00, 0.28] [0.00, 0.20] [0.01, 0.19] [0.01, 0.85] [0.01, 0.77] [0.00, 0.18] [0.01, 0.19] [0.01, 0.02] [0.01, 0.02] [0.01, 0.42] [0.02, 0.42] DKRL [0.03, 0.67] [0.01, 0.66] [0.00, 0.00] [0.00, 0.00] [0.26, 0.54] [0.27, 0.51] [0.00, 0.34] [0.00, 0.36] [0.43, 0.81] [0.44, 0.86] [0.00, 0.62] [0.00, 0.75] DARL [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.78, 0.92] [0.79, 0.93] [0.00, 0.02] [0.00, 0.02] [0.15, 0.70] [0.17, 0.74] [0.00, 0.00] [0.00, 0.00] DSH DHRL Joint Disease Sitting/ standing Range of primary memberships for features Physical disorders [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.57] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.07] [0.00, 0.18] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.23] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.15] [0.00, 0.00] [0.00, 0.00] [0.26, 0.80] [0.01, 0.62] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.16] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.35, 0.69] [0.54, 0.95] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.06, 0.53] [0.81, 0.93] [0.00, 0.00] [0.00, 0.00] [0.01, 0.34] [0.03, 0.53] [0.00, 0.00] [0.00, 0.00] AWESL AWESR AAKHL AAKHR ASHKL Table Calculated ranges of modified memberships and centre value for each physical disorder [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.92, 0.97] [0.11, 0.19] [0.00, 0.00] [0.00, 0.00] [0.01, 0.79] [0.04, 0.68] [0.00, 0.00] [0.00, 0.00] ASHKR [0.01, 0.11] [0.02, 0.48] [0.00, 0.00] [0.00, 0.00] [0.01, 0.52] [0.01, 0.51] [0.02, 0.74] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.03] [0.01, 0.63] [0.01, 0.46] [0.05, 0.69] [0.00, 0.00] [0.00, 0.00] [0.01, 0.54] [0.00, 0.12] [0.00, 0.00] [0.00, 0.16] [0.00, 0.00] [0.00, 0.00] [0.08, 0.54] [0.02, 0.48] AESSSL AESSSR [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.42, 0.86] [0.00, 0.15] [0.00, 0.00] [0.00, 0.00] [0.01, 0.71] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] AHSSSL [0.03, 0.49] [0.00, 0.00] [0.00, 0.56] [0.00, 0.00] [0.01, 0.43] [0.00, 0.00] [0.00, 0.17] [0.01, 0.50] [0.01, 0.68] [0.00, 0.01] [0.03, 0.18] [0.02, 0.51] AHSSSR (continued) [0.00,0.00] 0.00 [0.00,0.00] 0.00 [0.00,0.00] 0.00 [0.00,0.00] 0.00 [0.00,0.23] 0.115 [0.00,0.00] 0.00 [0.00,0.00] 0.00 [0.00,0.00] 0.00 [0.00,0.00] 0.00 [0.00,0.00] 0.00 [0.00,0.00] 0.00 [0.00,0.00] 0.00 Range [S c-min , S c-max ] after fuzzy Meet operation (centre Sc ) 122 S Saha et al Right Left Right Sitting Standing Standing [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.61, 0.77] [0.57, 0.78] [0.00, 0.40] [0.01, 0.42] [0.00, 0.34] [0.00, 0.40] [0.00, 0.00] [0.00, 0.00] [0.13, 0.70] [0.15, 0.66] [0.01, 0.71] [0.04, 0.78] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] DSH Left DARL Sitting DKRL Tennis elbow DERL DHRL Sitting/ standing Disease Joint Range of primary memberships for features Physical disorders Table (continued) [0.00, 0.00] [0.00, 0.16] [0.00, 0.00] [0.00, 0.14] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] AWESL AWESR AAKHL AAKHR ASHKL [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] ASHKR [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.01, 0.14] [0.00, 0.00] [0.00, 0.11] [0.00, 0.00] AESSSL AESSSR [0.00, 0.00] [0.00, 0.00] [0.00, 0.00] [0.02, 0.73] AHSSSL [0.01, 0.70] [0.00, 0.00] [0.00, 0.80] [0.00, 0.00] AHSSSR [0.00,0.00] 0.00 [0.00,0.00] 0.00 [0.00,0.00] 0.00 [0.00,0.00] 0.00 Range [S c-min , S c-max ] after fuzzy Meet operation (centre Sc ) Modified Type-2 Fuzzy Gesture Space Induced Physical Disorder Recognition 123 124 S Saha et al Performance Analysis The performance of the proposed modified T2FS-based physical disorder recognition algorithm is examined here with respect to three databases considered in this paper 5.1 Comparative Framework This paper compares the relative performance of the proposed modified T2FS algorithms with traditional gesture recognition algorithms/techniques The comparative framework includes traditional IT2FS [15], Support Vector Machine (SVM) classifier [17, 28], K-Nearest Neighbour (KNN) classification [16, 27], Levenberg– Marquardt Algorithm induced Neural Network (LMA-NN) [34], (Type-1) Fuzzy Relational Approach [5], Ensemble Decision Tree [35], and Radial Basis Function Network (RBFN) [14, 19] 5.2 Parameter Settings For all the contestant algorithms we used a common framework in terms of their features and databases We employ the best parametric set-up for all these algorithms as prescribed in their respective sources In our proposed method, the population-size of ABC is kept at 50 and the maximum function evaluations (FEs) is set as 100000 with ’limit’ of 100 In this paper, an asymmetric initialization procedure is adopted following the works reported in [8] 5.3 Performance Metric he physical disorder c of a subject (patient) in reality and that recognized by an algorithm may have four combinations as enlisted below True Positive of class c (TPc): the number of physical disorders that are recognized as in class c and indeed belong to class c False Positive of class c (FPc): the number of physical disorders that are recognized as in class c but in fact belong to class c ∈ [1, C] provided c = c False Negative of class c (FNc): the number of physical disorders that are recognized as in class c ∈ [1, C] provided c = c but indeed belong to class c True Negative of class c (TNc): the number of physical disorders that are correctly recognized to not to belong to class c In order to allow a quantitative assessment in performance of different gesture mediated physical disorder recognition algorithms, the comparative study uses the follow- Modified Type-2 Fuzzy Gesture Space Induced Physical Disorder Recognition 125 ing performance metrics to evaluate the performance of a multi-class classification algorithm with C classes (here C = 16) using the above four states and considering the micro-averaging strategy given in [14] (a) Recall: It measures the effectiveness of a classifier to identify the class labels Hence it relates to the test’s ability to identify positive results C TPc c=1 Recall = C (33) (TPc + FNc ) c=1 (b) Precision: It measures the degree of agreement of the data class labels with those of a classifier A high precision for a given test means that when the test yields a positive result, it is most likely correct in its assessment C TPc c=1 Precision = C (34) (TPc + FPc ) c=1 (c) Accuracy: It is the overall correctness of the predictive model and is calculated as the sum of correct predictions divided by the total number of predictions C Accuracy = c=1 TPc + TNc TPc + TNc + FPc + FNc (35) (d) F1_score: It is interpreted as a weighted average of the precision and recall, both of which are related to the performance of the positive instances An F1 score reaches its best value at and worst score at F1_score = (e) Average Error Rate: × Precision × Recall Precision + Recall (36) It is, in essence, the average per-class classification error C Average Error Rate = c=1 FPc + FNc TPc + TNc + FPc + FNc (37) 126 S Saha et al 5.4 Results and Performance Analysis The mean and the standard deviation (within parenthesis) of the best-of-run values of the performance metrics for 25 independent runs of each of the algorithm are presented in Table We also use paired t-test to compare the means of each performance metric produced by the best and the second best algorithm In the last column of Table we represent the statistical significance level of the difference of the means of the best two algorithms Note that here ’+’ indicates that the tvalue of 49 degrees of freedom is significant at a 0.05 level of significance by two-tailed test, whereas ” means the difference of means is not statistically significant, and ’NA’ stands for not applicable, covering cases for which two or more algorithms achieve the best accuracy results For all the t-tests carried in Table 6, the sample size is taken to be 25 The best algorithm is marked in bold A close inspection of Table indicates that the performance of the proposed modified T2FS-based physical disorder recognition algorithm has remained consistently superior to that of the other competitor methods It is interesting to see that out of performance metrics for all databases, i.e., out of 15 instances, in 12 cases, modified T2FS outperforms its nearest neighbour competitor in a statistically significant manner Here traditional IT2FS-based prediction method remains the second best algorithm 5.5 McNemar’s Statistical Test Let fA and fB be two classifiers obtained by algorithms A and B, when both the algorithms have a common training set R Let n01 be the number of examples misclassified by fA but not by fB , and n10 be the number of examples misclassified by fB but not by fA Then under the null hypothesis that both algorithms have the same error rate, the McNemar’s statistic Z in (38) follows a χ2 with degree of freedom equals to [9] Z= (|n01 − n10 | − 1)2 n01 + n10 (38) Let A be the proposed modified T2FS algorithm and B is one of the other seven algorithms We thus evaluate Z = Z1 through Z7 , where Zi denotes the comparator statistic of misclassification between the modified T2FS (Algorithm: A) and the i-th of the seven algorithms (Algorithm: B), where the suffix i refers to the algorithm in row number i of Table In Table the null hypothesis has been rejected, if Zi > χ21,α=0.05 = 3.84, where χ21,α=0.05 = 3.84 is the critical value of the chi square distribution for degree of freedom at probability of 0.05 [3] Performance metrics 0.8427 (0.136) 0.1690 (0.004) Fl_score Average error rate 0.8833 (0.011) 0.8298 (0.262) 0.8665 (0.078) 0.1802 (0.106) Precision Accuracy Fl_score Average error rate 0.8503 (0.196) 0.8410 (0.145) Accuracy Research fellows of Recall age 30–35 years 0.8808 (0.078) 0.8079 (0.165) Modified T2FS Algorithms Precision Research fellows of Recall age 25–30 years Database 0.1676 (0.004) 0.7949 (0.083) 0.8324 (0.162) 0.8415 (0.053) 0.7532 (0.243) 0.1759 (0.084) 0.8427 (0.144) 0.8241 (0.181) 0.8055 (0.083) 0.7946 (0.262) Traditional IT2FS 0.1704 (0.084) 0.7860 (0.152) 0.8296 (0.165) 0.8026 (0.075) 0.7701 (0.251) 0.1844 (0.106) 0.7717 (0.350) 0.8156 (0.259) 0.8023 (0.152) 0.7435 (0.450) SVM 0.1953 (0.259) 0.7636 (0.228) 0.8047 (0.311) 0.7998 (0.129) 0.7306 (0.254) 0.2065 (0.442) 0.7601 (0.549) 0.7935 (0.263) 0.7978 (0.228) 0.7259 (0.528) kNN Table Comparison of different physical disorder recognition algorithms for 25 Runs LMA-NN 0.2165 (0.399) 0.7343 (0.442) 0.7835 (0.528) 0.7742 (0.337) 0.6983 (0.349) 0.2248 (0.674) 0.7354 (0.579) 0.7752 (0.399) 0.7731 (0.53S) 0.7013 (0.601) fuzzy relation 0.2688 (0.774) 0.6824 (0.450) 0.7312 (0.601) 0.7615 (0.469) 0.6182 (0.473) 0.2415 (0.717) 0.7306 (0.622) 0.7585 (0.431) 0.7629 (0.825) 0.7010 (0.654) Ensemble decision tree 0.2822 (0.825) 0.6619 (0.538) 0.7178 (0.654) 0.6027 (0.530) 0.7340 (0.616) 0.2930 (0.768) 0.6418 (0.853) 0.7070 (0.500) 0.6027 (0.S33) 0.6865 (0.689) RBFN 0.2854 (0.817) 0.3381 (0.800) 0.7146 (0.689) 0.5858 (0.568) 0.7136 (0.814) 0.2992 (0.861) 0.6185 (0.869) 0.7008 (0.610) 0.5994 (0.876) 0.6390 (0.748) (continued) + + − + + + NA + + + Statistical significance Modified Type-2 Fuzzy Gesture Space Induced Physical Disorder Recognition 127 Performance metrics 0.8874 (0.099) 0.9348 (0.024) 0.8734 (0.196) 0.0952 (0.032) Accuracy Fl_score Average error rate 0.8600 (0.094) Modified T2FS Algorithms Precision Research fellows of Recall age 35–40 years Database Table (continued) Traditional IT2FS 0.1369 (0.312) 0.8212 (0.201) 0.8631 (0.072) 0.8444 (0.188) 0.7993 (0.157) SVM 0.1801 (0.336) 0.7932 (0.390) 0.8199 (0.223) 0.8121 (0.222) 0.7753 (0.219) kNN 0.1838 (0.530) 0.7666 (0.415) 0.8162 (0.415) 0.7999 (0.252) 0.7361 (0.347) LMA-NN 0.2153 (0.577) 0.7619 (0.442) 0.7847 (0.431) 0.7888 (0.270) 0.7368 (0.373) fuzzy relation 0.2582 (0.752) 0.7413 (0.464) 0.7418 (0.620) 0.7665 (0.360) 0.7178 (0.505) Ensemble decision tree 0.2714 (0.776) 0.6475 (0.491) 0.7286 (0.680) 0.6032 (0.512) 0.6989 (0.769) RBFN 0.2744 (0.839) 0.6422 (0.818) 0.7256 (0.789) 0.6367 (0.604) 0.6479 (0.832) + + + + + Statistical significance 128 S Saha et al 72 80 81 34 32 28 LMA-NN Fuzzy Relation 31 31 kNN Ensemble Decision Tree RBFN 84 62 57 36 SVM 55 37 Traditional IT2FS 27.0089 21.4375 20.7567 14.6250 7.5937 4.3011 3.1413 Reject Reject Reject Reject Reject Reject Accept 35 37 38 40 41 41 43 107 103 99 84 69 65 61 n10 35.5000 30.1786 26.2774 14.9113 6.6273 4.9905 2.7788 Zi Reject Reject Reject Reject Reject Reject Accept Comment on acceptance/rejection of null hypothesis n01 Comment on acceptance/rejection of null hypothesis n01 Zi research scholars of ase 30–35 research scholars of aee 25–30 n10 Jadavpur University database of Jadavpur University database of Competitor Control algeorithm A = modified T2FS algorithm = B Table Performance analysis using McNemar’s test 26 26 26 27 29 30 32 n01 86 75 70 65 63 58 55 n10 31.0804 22.8119 19.2604 14.8804 11.8371 8.2841 5.5632 Zi Reject Reject Reject Reject Reject Reject Reject Comment on acceptance/rejection of null hypothesis research scholars of aee 35–40 Jadavpur University database of Modified Type-2 Fuzzy Gesture Space Induced Physical Disorder Recognition 129 130 S Saha et al 5.6 Friedman Test A non-parametrical statistical test, known as Friedman two-way analysis of variances by ranks [11], is also performed on the mean of Accuracy metric for 25 independent runs of each of the eight algorithms as reported in Table for all three databases Let j ri be the ranking of the observed Accuracy obtained by the i-th algorithm for the j-th database The best of all the k algorithms, i.e., i = [1, k], is assigned a rank of and the worst is assigned the ranking k Table summarizes the rankings obtained by Friedman procedure Then the average ranking acquired by the i-th algorithm over all j = [1, N ] algorithms is defined as follows Ri = N N j ri (39) j=1 Under the null hypothesis, formed from the supposition that the results of the algorithms are equivalent and, therefore, their rankings are also similar, the Friedman’s statistic in (40) follows a χ2F distribution with degree of freedom equals to k − [11] K 12N k(k + 1)2 (40) χ2F = R2i − k(k + 1) i=1 In this paper N = number of databases considered = and k = number of competitor algorithms = In Table 8, it is shown that the null hypothesis has been rejected, as χ2F = 20.7779 is greater than χ27,α=0.05 = 14.067, where χ27,α=0.05 = 14.067 is the critical value of the χ2F distribution for k − = degrees of freedom at probability of 0.05 [40] 5.7 Iman-Davenport Statistical Test Additionally, we use Iman-Davenport test as a variant of Friedman test that provides better statistics [11] It is a derivation from the Friedman’s statistic given that the metric in (40) produces a conservative undesirably effect The statistic proposed by Iman-Davenport is given as follows FF = (N − 1) × χ2F N × (k − 1) − χ2F (41) It is distributed according to F distribution with (k − 1) and (k − 1) × (N − 1) degrees of freedom For k = and N = 3, it is shown in Table that the null hypothesis has been rejected, as FF = 187.0946 is greater than F , 14, α = 0.05 = j 2.606 Traditional IT2FS SVM KNN LMA-NN Fuzzy Relation Ensemble decision tree RBFN Critical difference for α = 0.05 25–30 (ri1 ) 30–35 (ri2 ) 35–40 (ri3 ) Jadavpur University database of research scholars of age Ranking ri of the i-th algorithm obtained through j-th Modified T2FS Algorithm 8.000 7.000 6.000 5.000 4.000 3.000 1.667 1.333 Average rankins Ri Table Performance analysis using Friedman and Iman-Davenport tests 20.7779 Statistics χF Reject Comment on acceptance/rejection of null hypothesis Friedman test 187.0946 Statistics FF Reject Comment on acceptance/rejection of null hypothesis Iman-Davenport test Modified Type-2 Fuzzy Gesture Space Induced Physical Disorder Recognition 131 132 S Saha et al 3.38, where F , 14, α = 0.05 = 3.38 is the critical value of the F distribution for k − = and (k − 1) × (N − 1) = 14 degrees of freedom at probability of 0.05 [40] 5.8 Bonferroni-Dunn Test The results in Table highlight modified T2FS as the best algorithm, so the post-hoc analysis [32] is performed with modified T2FS as the control method In the post-hoc analysis we apply the Bonferroni-Dunn test [32] over the results of Friedman procedure The analysis indicates the level of significance at which the control algorithm is better than each of the remaining algorithms (i.e., when the null hypothesis is rejected) For the Bonferroni-Dunn test, a critical difference (CD) [40] is calculated which for these data comes as 2.606 The interpretation of this measure is that the performance of two algorithms is significantly different, only if the corresponding average ranks differ by at least a critical difference, which is depicted in Fig 15 It can be perceived that only for Traditional IT2FS and SVM, the null hypothesis cannot be rejected with any of the tests for the level of significance α = 0.05 The other five algorithms, however, are regarded as significantly worse than the modified T2FS Fig 15 Graphical representation of Bonferroni-Dunn’s procedure considering modified T2FS as control method Modified Type-2 Fuzzy Gesture Space Induced Physical Disorder Recognition 133 Conclusion The paper presents an automatic physical disorder recognition system by utilizing the composite benefit of GT2FS and IT2FS In order to classify an unknown gestural expression, the system makes use of the background knowledge about a large gesture database with known physical disorder classes The modified T2FS-based recognition scheme, which is a combined version of GT2FS and IT2FS, requires type-2 secondary membership functions, which are obtained using an innovative evolutionary approach that is also proposed in this paper The proposed strategy first constructs a type-2 fuzzy gesture space, and then infers the physical disorder class of the unknown gestural expression by determining the maximum support of the individual physical disorder classes using the pre-constructed fuzzy gesture space The class with the highest support is assigned as the physical disorder of the unknown gestural expression The traditional IT2FS-based recognition scheme takes care of the inter-subject level uncertainty in computing the maximum support of individual physical disorder class However, the modified T2FS-based recognition scheme, with the help of GT2FS, takes care of both the inter- and intra-subject level uncertainty, and thus offers higher classification accuracy for the same set of features Using three data sets, the classification accuracy obtained by employing modified IT2FS is 93.48%, and by IT2FS is 86.31% Acknowledgements The funding of University Grant Commission, University of Potential for Excellence Program (Phase II) in Cognitive Science, Jadavpur University, India are greatly acknowledged for the present work References Acampora, Giovanni, Daniyal Alghazzawi, Hani Hagras, and Autilia Vitiello 2016 An interval type-2 fuzzy logic based framework for reputation management in peer-to-peer e-commerce Information Sciences 333: 88–107 Basturk, Bahriye 2006 An artificial bee colony (abc) algorithm for numeric function optimization In IEEE Swarm Intelligence Symposium, 2006, Indianapolis, IN, USA Bickel, P.J., and K.A Doksum 1977 Mathematical Statistics Holden-Day SF: Inc Burba, Nathan, Mark Bolas, David M Krum, and Evan A Suma 2012 Unobtrusive measurement of subtle nonverbal behaviors with the microsoft kinect In 2012 IEEE Virtual reality short papers and posters (VRW), 1–4 IEEE Chakraborty, Aruna, Amit Konar, Uday Kumar Chakraborty, and Amita Chatterjee 2009 Emotion recognition from facial expressions and its control using fuzzy logic IEEE Transactions on Systems, Man, and Cybernetics-Part A: Systems and Humans 39 (4): 726–743 Chen, Chen-Tung and Wei-Zhan Hung 2017 Evaluating the service quality of hospital by using topsis with interval type-2 fuzzy sets In 2017 International Conference on Fuzzy Theory and its Applications (iFUZZY), 1–5 IEEE Clark, Ross A., Yong-Hao Pua, Karine Fortin, Callan Ritchie, Kate E Webster, Linda Denehy, and Adam L Bryant 2012 Validity of the microsoft kinect for assessment of postural control Gait & Posture 36 (3): 372–377 134 S Saha et al Das, Swagatam, Ajith Abraham, Uday K Chakraborty, and Amit Konar 2009 Differential evolution using a neighborhood-based mutation operator IEEE Transactions on Evolutionary Computation 13 (3): 526–553 Dietterich, Thomas G 1998 Approximate statistical tests for comparing supervised classification learning algorithms Neural Computation 10 (7): 1895–1923 10 Galna, Brook, Gillian Barry, Dan Jackson, Dadirayi Mhiripiri, Patrick Olivier, and Lynn Rochester 2014 Accuracy of the microsoft kinect sensor for measuring movement in people with parkinson’s disease Gait & Posture 39 (4): 1062–1068 11 García, Salvador, Daniel Molina, Manuel Lozano, and Francisco Herrera 2009 A study on the use of non-parametric tests for analyzing the evolutionary algorithms’ behaviour: a case study on the cec’2005 special session on real parameter optimization Journal of Heuristics 15 (6): 617 12 Halder, Anisha, Amit Konar, Rajshree Mandal, Aruna Chakraborty, Pavel Bhowmik, Nikhil R Pal, and Atulya K Nagar 2013 General and interval type-2 fuzzy face-space approach to emotion recognition IEEE Transactions on Systems, Man, and Cybernetics: Systems 43 (3): 587–605 13 Halder, Anisha, Pratyusha Rakshit, Sumantra Chakraborty, Amit Konar, Aruna Chakraborty, Eunjin Kim, and Atulya K Nagar 2012 Reducing uncertainty in interval type-2 fuzzy sets for qualitative improvement in emotion recognition from facial expressions In 2012 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), 1–8 IEEE 14 Kim, Dae-Jin, Sang-Wan Lee, and Zeungnam Bien 2005 Facial emotional expression recognition with soft computing techniques In The 14th IEEE International Conference on Fuzzy Systems, 2005 FUZZ’05, 737–742 IEEE 15 Klir, George, and Bo Yuan 1995 Fuzzy Sets and Fuzzy Logic, vol NJ: Prentice Hall 16 Lai, Kam, Janusz Konrad, and Prakash Ishwar 2012 A gesture-driven computer interface using kinect In 2012 IEEE Southwest Symposium on Image Analysis and Interpretation (SSIAI), 185–188 IEEE 17 Le, Thi-Lan, Minh-Quoc Nguyen, et al 2013 Human posture recognition using human skeleton provided by kinect In 2013 International Conference on Computing, Management and Telecommunications (ComManTel), 340–345 IEEE 18 Lee, Ching-Hung, Jang-Lee Hong, Yu-Ching Lin, and Wei-Yu Lai 2003 Type-2 fuzzy neural network systems and learning International Journal of Computational Cognition (4): 79–90 19 Lisetti, Christine L., and David E Rumelhart 1998 Facial expression recognition using a neural network In FLAIRS Conference, 328–332 20 Martin, Chris C., Dan C Burkert, Kyung R Choi, Nick B Wieczorek, Patrick M McGregor, Richard A Herrmann, and Peter A Beling 2012 A real-time ergonomic monitoring system using the microsoft kinect In 2012 IEEE Systems and Information Design Symposium (SIEDS), 50–55 IEEE 21 Mendel, Jerry M 2001 On the importance of interval sets in type-2 fuzzy logic systems In Joint 9th IFSA World Congress and 20th NAFIPS International Conference, 2001, vol 3, 1647–1652 IEEE 22 Mendel, Jerry M., RI Bob John 2002 Type-2 fuzzy sets made simple IEEE Transactions on Fuzzy Systems 10 (2): 117–127 23 Mendel, Jerry M., Robert I John, and Feilong Liu 2006 Interval type-2 fuzzy logic systems made simple IEEE Transactions on Fuzzy Systems 14 (6): 808–821 24 Metcalf, Cheryl D, Rebecca Robinson, Adam J Malpass, Tristan P Bogle, Thomas A Dell, Chris Harris, and Sara H Demain 2013 Markerless motion capture and measurement of hand kinematics: validation and application to home-based upper limb rehabilitation IEEE Transactions on Biomedical Engineering 60 (8): 2184–2192 25 Nixon, Mark, and Alberto S Aguado 2012 Feature Extraction and Image Processing for Computer Vision Academic Press 26 Olivas, Frumen, Fevrier Valdez, Oscar Castillo, Claudia I Gonzalez, Gabriela Martinez, and Patricia Melin 2017 Ant colony optimization with dynamic parameter adaptation based on interval type-2 fuzzy logic systems Applied Soft Computing 53: 74–87 Modified Type-2 Fuzzy Gesture Space Induced Physical Disorder Recognition 135 27 Oszust, Mariusz, and Marian Wysocki 2013 Recognition of signed expressions observed by kinect sensor In 2013 10th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS), 220–225 IEEE 28 Parajuli, Monish, Dat Tran, Wanli Ma, and Dharmendra Sharma 2012 Senior health monitoring using kinect In 2012 Fourth International Conference on Communications and Electronics (ICCE), 309–312 IEEE 29 Quante, Markus, Stephanie Hille, Markus D Schofer, Jürgen Lorenz, and Michael Hauck 2008 Noxious counterirritation in patients with advanced osteoarthritis of the knee reduces mcc but not sii pain generators: A combined use of meg and eeg Journal of Pain Research 1:1 30 Rakshit, Pratyusha, Aruna Chakraborty, Amit Konar, and Atulya K Nagar 2013 Secondary membership evaluation in generalized type-2 fuzzy sets by evolutionary optimization algorithm In 2013 IEEE International Conference on Fuzzy Systems (FUZZ), 1–8 IEEE 31 Rakshit, Pratyusha, Amit Konar, Archana Chowdhury, Eunjin Kim, and Atulya K Nagar 2013 Muti-objective evolutionary approach of ligand design for protein-ligand docking problem In 2013 IEEE Congress on Evolutionary Computation (CEC), 237–244 IEEE 32 Rakshit, Pratyusha, Amit Konar, Swagatam Das, Lakhmi C Jain, and Atulya K Nagar 2014 Uncertainty management in differential evolution induced multiobjective optimization in presence of measurement noise IEEE Transactions on Systems, Man, and Cybernetics: Systems 44 (7): 922–937 33 Rakshit, Pratyusha, Amit Konar, Swagatam Das, and Atulya K Nagar 2013 Abc-tdql: An adaptive memetic algorithm In Hybrid Intelligent Models and Applications (HIMA), 35–42 IEEE 34 Saha, Sriparna, Monalisa Pal, Amit Konar, and Ramadoss Janarthanan 2013 Neural network based gesture recognition for elderly health care using kinect sensor In International Conference on Swarm, Evolutionary, and Memetic Computing, 376–386 Springer 35 Stone, Erik E., and Marjorie Skubic 2015 Fall detection in homes of older adults using the microsoft kinect IEEE Journal of Biomedical and Health Informatics 19 (1): 290–301 36 Tong, Jing, Jin Zhou, Ligang Liu, Zhigeng Pan, and Hao Yan 2012 Scanning 3d full human bodies using kinects IEEE Transactions on Visualization and Computer Graphics 18 (4): 643–650 37 Wagner, Christian, Hani Hagras, et al 2010 Toward general type-2 fuzzy logic systems based on zslices IEEE Transactions on Fuzzy Systems 18 (4): 637–660 38 Xiaoyi Yu, Lingyi Wu, Qingfeng Liu, and Han Zhou 2011 Children tantrum behaviour analysis based on kinect sensor In 2011 Third Chinese Conference on Intelligent Visual Surveillance (IVS) 49–52 IEEE 39 Zadeh, Lotfi A., et al 1965 Fuzzy sets Information and Control (3): 338–353 40 Zar, Jerrold H., et al 1999 Biostatistical Analysis Pearson Education India 41 Zhai, Daoyuan, and Jerry M Mendel 2011 Uncertainty measures for general type-2 fuzzy sets Information Sciences 181 (3): 503–518 42 Zhang, Zhengyou 2012 Microsoft kinect sensor and its effect IEEE Multimedia 19 (2): 4–10 43 Zimmermann H.-J 1996 Fuzzy control In Fuzzy Set Theory–and Its Applications, 203–240 Springer ... Acampora Witold Pedrycz Athanasios V Vasilakos Autilia Vitiello • • • Editors Computational Intelligence for Semantic Knowledge Management New Perspectives for Designing and Organizing Information Systems... Technologies, University of Naples “Parthenope , Naples, Italy Autilia Vitiello University of Naples Federico II, Naples, Italy Raciel Yera University of Ciego de Ávila, Ciego de Ávila, Cuba Acronyms... University of Jắn, Jắn, Spain e-mail: martin@ujaen.es © Springer Nature Switzerland AG 2020 G Acampora et al (eds. ), Computational Intelligence for Semantic Knowledge Management, Studies in Computational
- Xem thêm -

Xem thêm: Computational intelligence for semantic knowledge management, 1st ed , giovanni acampora, witold pedrycz, athanasios v vasilakos, autilia vitiello, 2020 313 , Computational intelligence for semantic knowledge management, 1st ed , giovanni acampora, witold pedrycz, athanasios v vasilakos, autilia vitiello, 2020 313

Mục lục

Xem thêm

Gợi ý tài liệu liên quan cho bạn