1,721,548 research outputs found
Delaying Inconsistency Resolution Using Fuzzy Logic
No full textWhile developing complex systems, software engineers
generally have to deal with various kinds of inconsistencies. Certain kinds of inconsistencies are inevitable, for
instance, in case of multiple persons working independently of each other within the same project. Some inconsistencies are desirable when, for instance, alternative
solutions exist for the same problem, and these solutions
have to be preserved to allow further refinements along
the development process. Current software development
methods do not provide adequate means to model the desired inconsistencies and, therefore, aim to resolve the
inconsistencies whenever they are detected. Although
early resolution of inconsistencies reduces complexity of
design by eliminating possible alternatives, it results in
loss of information and excessive restriction of the design
space. This paper aims to enhance the current methods by
modelling and controlling the desired inconsistencies
through the application of fuzzy logic
Feature Selection based on a Modified Fuzzy C-means Algorithm with Supervision
No full textIn this paper we propose a new approach to feature selection based on a modified fuzzy C-means algorithm with supervision (MFCMS). MFCMS completes the unsupervised learning of classical fuzzy C-means with labeled patterns. The labeled patterns allow MFCMS to accurately model the shape of each cluster and consequently to highlight the features which result to be particularly effective to characterize a cluster. These features are distinguished by a low variance of their values for the patterns with a high membership degree to the cluster. If, with respect to these features, the distance between the prototype of the cluster and the prototypes of the other clusters is high, then these features have the property of discriminating between the cluster and the other clusters. To take these two aspects into account, for each cluster and each feature, we introduce a purposely defined index: the higher the value of the index, the higher the discrimination capability of the feature for the cluster. We execute MFCMS on the training set considering all patterns as labeled. Then, we retain the features which are associated, at least for one cluster, with an index larger than a threshold τ. We applied MFCMS to several real-world pattern classification benchmarks. We used the well-known k-nearest neighbors as learning algorithm. We show that feature selection performed by MFCMS achieved an improvement in generalization on all data sets
Recognition of olfactory signals based on supervised fuzzy c-means and k-NN algorithms
No full textIn this paper we present a novel method for odour recognition based on a supervised fuzzy C-means (SFCM) algorithm and a k-nearest neighbour (k-NN) algorithm. The method is applied to experimental data collected from a sensor array composed of metal oxide sensors (MOSs). The sensors are exposed to odourants and the relative resistance values are used for classification. SFCM selects the features, which better characterize the sensor responses, and computes both the memberships of the odourants to classes, and the shape of classes. k-NN exploits the output of SFCM to recognize unknown odourants. We describe the application of the method to the classification of food packages and show that the experimental results support the methodology presented
Fuzzy Logic Based Object-Oriented Methods to Reduce Quantization Error and Contextual Bias Problems in Software Development
No full textDuring the last several years, a considerable number of software development methods have been introduced to produce robust, reusable and adaptable software systems. Methods create software artifacts through the application of a large number of heuristic rules. These rules are generally expressed in two-valued logic. In object-oriented methods, for instance, candidate classes are identified by applying the following intuitive rule: "If an entity in a requirement specification is relevant and can exist autonomously in the application domain, then select it as a class". In this paper, we identify and define two major problems regarding how rules are defined and applied in current methods. First, two-valued logic cannot effectively express the approximate and inexact nature of a typical software development process. Although software engineers can perceive partial relevance of an entity and possibly select the entity as a partial candidate class, they are constrained by two-valued logic to quantize relevance into relevant and irrelevant. Second, the influence of contextual factors on rules is generally not modelled explicitly. We term these problems as quantization error and contextual bias problems, respectively. To reduce these problems, we propose to express heuristic rules using fuzzy logic. We illustrate formally how fuzzy logic-based methodological rules can help in lowering the effects of quantization error and contextual bias problems
Feature Selection based on Similarity
No full textA new method for feature selection is proposed. The method associates a weight with each feature by minimising an appropriate index defined in terms of similarity between patterns of the training set. The weight measures the importance of the corresponding feature in characterising the classes. Features associated with low weights are considered irrelevant and therefore eliminated. Experimental results to confirm the validity of the method are shown
Fuzzy classification of handwritten characters
No full textIn this paper we present EYE, a fuzzy logic-based classifier for off-line recognition of isolated handwritten characters. EYE uses a new linguistic classification method based on the linguistic description of the shape of the character. The linguistic terms are derived from a fuzzy partition of the space occupied by the character EYE classifies characters written by any writer. A small scale application of the method in which 26 lower-case cursive characters written by 30 and 20 different writers were used as training and test sets, respectively, yielded 68.2% recognition rate
- …
