Session Program


  • 10 July 2017
  • 08:00AM - 10:00AM
  • Room: Auditorium
  • Chairs: José Antonio Sanz and Mikel Galar

Fuzzy Methods and Data Mining I: Rules Learning

Abstract - This paper proposes a multilabel fuzzy decision tree classifier named MLFuzzDT. The algorithm uses generalized fuzzy entropy, aggregated over all labels, to choose the best attribute for growing the tree. The proposed algorithm also can generate leaves predicting partial label sets, which can incorporate to some degree the dependence among labels, as well as produce more interpretable models. An empirical analysis shows that, although the algorithm does not yet incorporate pruning nor fuzzy interval adjustment phases, it is competitive with other tree based approaches for multilabel classification, with better performance in data sets having numerical features that can be fuzzified.
Abstract - Boosting is a simple and effective procedure that combines several weak learners with the aim of generating a strong classifier. Multi-class boosting has been only recently studied in the context of crisp classifiers, showing encouraging performances. In this paper, we propose FDT-Boost, a boosting approach shaped according to the multi-class SAMME-AdaBoost scheme, that employs size-constrained fuzzy binary decision trees as weak classifiers. We test FDT-Boost on twenty-three classification benchmarks. By comparing our approach with FURIA, one of the most popular fuzzy classifiers, and with a fuzzy binary decision tree, we show that our approach is accurate, yet keeping low the model complexity in terms of total number of leaf nodes.
Abstract - In a recent paper, it was introduced the concept of Choquet-like Copula- based integral (CC-integral for short). This kind of function extends the standard Choquet integral and generalizes it by copula functions. These functions were applied in the Fuzzy Reasoning Method (FRM) of a Fuzzy Rule- Based Classification System (FRBCS), presenting an example where the CC- integral based on the minimum t-norm had different behaviors according to the values being aggregated. Therefore, the resulting FRM is theoretically more flexible than those associated with classical aggregation functions like the maximum. In this work, we present a methodology to study the flexibility of the aggregation function used in the FRM. Specifically, we conduct an analysis of 3 different methods to aggregate values in the FRM, namely, the CC- integral based on the minimum t-norm, the standard Choquet integral and the maximum (classical FRM of the winning rule -- WR). We prove that the CC- integral behaves in different ways according to the values to be aggregated, whereas the Choquet integral offers an averaging behavior and the WR presents an strict behavior, since it considers only the rule having the maximum compatibility with the example.
Abstract - Identification of presence of target acoustic sound or event from the single channel mixture is a challenging task of automatic sound recognition system. In presence of background detection and classification of target acoustic event becomes more difficult.Various methods have been proposed that extract features from spectrogram of sound and then extracted features are used with traditional non negative matrix factorization for separation of overlapping sound. In this paper, we propose an approach to separate and classify single channel acoustic events. The method combines Common Fate Transformation and Cauchy Non-negative Matrix Factorization for feature extraction and finally fuzzy rule based classifier is developed for classification. The proposed method, when applied to real data, gave high true positive rate. The method also gave better results in terms of true positive rate when compared to widely used support vector machine using same real data. Moreover, the proposed approach is fast and can be used for the efficient separation of acoustic events from overlapping sounds.
Abstract - In real life, most of information is presented with words. Also, classification is an important issue to make decisions in daily life. Fuzzy logic gives flexibility to handle the imprecise information for computing with words. And, linguistic variables can be defined by using triangular fuzzy numbers given as L-R fuzzy numbers. This study aims to provide a classification approach by using fuzzy ID3 algorithm for linguistic data. In this study, Weighted Averaging Based on Levels (WABL) method, fuzzy c-means, and fuzzy ID3 algorithm are combined. WABL method is used to obtain crisp data set. Then, fuzzy c-means (FCM) algorithm is performed to introduce the shape of the linguistic terms limits and to obtain each membership degree of each linguistic term defined for fuzzy variables in data sets. At last, Fuzzy ID3 algorithm is applied. The rules are generated and the reasoning is done by using Zadeh T-norm/conorm operators. Experimental study is performed on six well-known data sets (Iris, Wdcb, Phoneme, Ring, Sonar, and Pima). As a conclusion, we proposed a fuzzy decision tree classification methodology for linguistic datasets.