Karim Saheb Ettabaa

Ontology Knowledge Mining Based Association Rules Ranking

Medical association rules induction is used to discover useful correlations between pertinent concepts from large medical databases. Nevertheless, ARs algorithms produce huge amount of delivered rules and do not guarantee the usefulness and interestingness of the generated knowledge. To overcome this drawback, we propose an ontology based interestingness measure for ARs ranking. According to domain expert, the goal of the use of ARs is to discover implicit relationships between items of different categories such as clinical features and disorders,clinical features and radiological observations, etc. Thats to say, the itemsets which are composed of similar items are uninteresting. Therefore, the dissimilarity between the rules items can be used to judge the interestingness of association rules; the more different are the items, the more interesting the rule is. In this paper, we design a distinct approach for ranking semantically interesting association rules involving the use of an ontology knowledge mining approach. The basic idea is to organize the ontologys concepts into a hierarchical structure of conceptual clusters of targeted subjects, where each cluster encapsulates similar concepts suggesting a specific category of the domain knowledge. The interestingness of association rules is, then, defined as the dissimilarity between corresponding clusters. Thats to say, the further are the clusters of the items in the AR, the more interesting the rule is. We apply the method in our domain of interest - mammographic domain-using an existing mammographic ontology called Mammo*, with the goal of deriving interesting rules from past experiences, to discover implicit relationships between concepts modeling the domain.*http://sourceforge.net/p/gimimammography/code/HEAD/tree/trunk/owl

Fuzzy Clustering based Approach for Ontology Alignment

Recently, several ontologies have been proposed for real life domains, where these propositions are large and voluminous due to the complexity of the domain. Consequently, Ontology Aligning has been attracting a great deal of interest in order to establish interoperability between heterogeneous applications. Although, this research has been addressed, most of existing approaches do not well capture suitable correspondences when the size and structure vary vastly across ontologies. Addressing this issue, we propose in this paper a fuzzy clustering based alignment approach which consists on improving the ontological structure organization. The basic idea is to perform the fuzzy clustering technique over the ontologyâx80x99s concepts in order to create clusters of similar concepts with estimation of medoids and membership degrees. The uncertainty is due to the fact that a concept has multiple attributes so to be assigned to different classes simultaneously. Then, the ontologies are aligned based on the generated fuzzy clusters with the use of different similarity techniques to discover correspondences between conceptual entities.

Ontology Knowledge Mining for Ontology Alignment

AbstractAs the ontology alignment facilitates the knowledge exchange among the heterogeneous data sources, several methods have been introduced in literature. Nevertheless, few of them have been interested in decreasing the problem complexity and reducing the research space of correspondences between the input ontologies.This paper presents a new approach for ontology alignment based on the ontology knowledge mining. The latter consists on producing for each ontology a hierarchical structure of fuzzy conceptual clusters, where a concept can belong to several clusters simultaneously. Each level of the hierarchy reflects the knowledge granularity degree of the knowledge base in order to improve the effectiveness and speediness of the information retrieval. Actually, such method allows the knowledge granularity analyze between the ontologies and facilitates several ontology engineering techniques. The ontology alignment process is performed iteratively over the produced hierarchical structure of the fuzzy cl...

Adaptive Batch Extraction for Hyperspectral Image Classification Based on Convolutional Neural Network

Deep Learning for Hyperspectral Imaging Classification is a wonderful solution, despite a few fuzzification. Conventional neural networks are very effective for classification tasks which have allowed them to be used by a very large companies. In this paper, we present an approach to initialize the convolutional data: Firstly, an adaptive selection of kernels by a clustering algorithm; Secondly, by the definition of adaptive batches size. In order to validate our proposed approach, we tested the algorithms on three different hyperspectral images, and the results showed the effectiveness of our proposal.

An unsupervised classification approach for hyperspectral images based adaptive spatial and spectral neighborhood selection and graph clustering

Abstract In remote sensing image processing, the classification is an interesting step to distinguish the image scene composition and can be of interesting role in different applications such as environmental monitoring and geological studies. Unlike the clustering, the classification needs labeled data for the training; however gaining these labeled data was always an expensive and hard task. For that, in this paper we propose an unsupervised classification approach that gains its labeled data from the proposed spatial and spectral graph clustering approach. The proposed adaptive spatial and spectral neighborhood selection approach is an extension of the k nearest neighborhood that assigns an adaptive number of neighbors to each pixel depending in its spatial and spectral relationship to its neighboring pixels. Then, this neighborhood will be clustered, to provide the first labeled training set, based on a hierarchical graph clustering algorithm. Finally, an SVM classification with a recursive kernel will be performed on the selected first labeled data at a first step and then the classification results are improved with the classification iterations of the recursive kernel. Experimental results on real hyperspectral images proved that with few iterations of the recursive kernel the proposed approach results are similar and even better then the supervised classification

Fast Spatial Spectral Schroedinger Eigenmaps algorithm for hyperspectral feature extraction

Abstract Based on the Laplacian Eigenmaps (LE) algorithm and a potential matrix, the Spatial Spectral Schroedinger Eigenmaps (SSSE) technique has proved a great yield during the hyperspectral dimensionality reduction process. Experimentally, SSSE is in deficiency of high computing time which may hinder its contribution in the remote sensing field. In this paper, a fast variant of the SSSE approach, called Fast SSSE, was proposed. The new suggested method substitutes the quadratic constraint employed during the optimization problem, by a linear constraint. This overhaul preserves the data properties in analogous way to the SSSE technique, but with a fast implementation. Two real hyperspectral data sets were adopted during the experimental process. Experiment analysis exhibited good classification accuracy with a reduced computational effort, compared with the original SSSE approach.

Hyperspectral image betweenness centrality clustering based adaptive spatial and spectral neighborhood approach for anomaly detection

Segmentation-based anomaly detectors proceeds to the clustering of the hyperspectral image as a first step. However, most of the well-known clustering methods cluster anomalous pixels as a part of the background. This paper presents a new hyperspectral image clustering approach based on the betweenness centrality measure. The proposed approach starts by the construction of an adaptive spatial and spectral neighborhood for each pixel. This neighborhood is based on the selection of the nearest spectral and spatial neighbors in multiple windows around each pixel to allow well-suited representation of the image features. In the next step, this neighborhood is clustered based on the edge betweenness measure algorithm that splits the image into regions sharing similar features. This approach (1) allows the reduction of intercluster relationship, (2) favors intracluster relations, and (3) preserves small clusters that can hold anomalous pixels. Experimental results show that the proposed approach is efficient for clustering and overcomes the state of the art approaches.

Anomaly detection in hyperspectral images based spatial spectral classification

Anomaly detection in hyperspectral images aims at detecting small size objects of unknown spectra. The major problem with anomaly detection is the absence of prior knowledge. Consequently, the extraction of true anomalies from the background and noise is a challenging task. In fact, the image scene already contains the background, noises and anomalous pixels and even in presence of prior knowledge, the differentiation between these contents is often challenging and can lead to a high false alarm rate. In this paper, a new approach for anomaly detection is proposed. The approach aims at generating knowledge about the scene before anomaly detection. This knowledge is derived from a semi-supervised SVM classification based on the betweenness centrality clustering of the spatial and spectral graph of the image. Anomaly detection is performed then, based on the Mahalanobis distance between different classes of the image. Our experimental results show improvement in the detection rate compared to the benchmark anomaly detectors.

Hyperspectral image feature selection for the fuzzy c-means spatial and spectral clustering

Hyperspectral image clustering is commonly applied for unsupervised classification. However, the clustering results of traditional methods are not sufficient seeing the nature of the image as a data cube with high dimensionality. In addition, the complex relations between spatial neighboring pixels are not considered in traditional methods. In this paper the fuzzy c-means clustering is revisited and customized. The proposed approach aims at the reduction of dimensionality of the data cube while preserving the most relevant spectral features and the improvement of the clustering result. The integration of spatial feature can express natural dependence between neighboring pixels and enhance the clustering. For that the presented approach starts by a band selection method based on the hierarchical clustering of spectral bands using the mutual information measure to reduce the dimensionality of the image. Then, a new version of the fuzzy c-means clustering algorithm is proposed; this version includes spatial and spectral features. Experimental result on real hyperspectral data shows an improvement on the accuracy over conventional clustering methods.