1 September 2021 in International Journal of Computational Intelligence Systems
Diabetes is an extremely serious hazard to global health and its incidence is increasing vividly. In this paper, we develop an effective system to diagnose diabetes disease using a hybrid optimization-based Support Vector Machine (SVM).The proposed hybrid optimization technique integrates a Crow Search algorithm (CSA) and Binary Grey Wolf Optimizer (BGWO) for exploiting the full potential of SVM in the diabetes diagnosis system. The effectiveness of our proposed hybrid optimization-based SVM (hereafter called CS-BGWO-SVM) approach is carefully studied on the real-world databases such as UCIPima Indian standard dataset and the diabetes type dataset from the Data World repository. To evaluate the CS-BGWO-SVM technique, its performance is related to several state-of-the-arts approaches using SVM with respect to predictive accuracy, Intersection Over-Union (IoU), specificity, sensitivity, and the area under receiver operator characteristic curve (AUC). The outcomes of empirical analysis illustrate that CS-BGWO-SVM can be considered as a more efficient approach with outstanding classification accuracy. Furthermore, we perform the Wilcoxon statistical test to decide whether the proposed cohesive CS-BGWO-SVM approach offers a substantial enhancement in terms of performance measures or not. Consequently, we can conclude that CS-BGWO-SVM is the better diabetes diagnostic model as compared to modern diagnosis methods previously reported in the literature.
29 June 2017 in International Journal of Fuzzy Systems
1 November 2019 in Journal of Physics: Conference Series
Hadoop is a widely used analytical tool for analyzing large volume of data which are in terms of Exa & Zetta bytes. Hadoop Distributed File System(HDFS) is a well-known storage system which provides distributed storage using commodity hardware. Since Big Data consists of highly sensitive data like Electronic Health Records, Financial Data etc., harnessing big data access and security plays a vital role. The bottom most layer of security can be provided by encryption. Since the data stored is highly unstructured, wider in variety and the velocity at which data is stored is also unpredictable, traditionally used ciphers like AES, RSA, RC4 etc. will not work efficiently on Big Data. The solution must have to consider at-risk sensitive data, variety of data which needs intelligence to classify the data. When data blocks are stored in data nodes, with each block 4-bits called sensitivity bits has to be added. Two bits indicate the level of sensitivity and two more bits indicate the type of data. These 4-bits are used by the map and reduce functions to effectively encrypt the required data blocks alone and not all, in a time efficient manner by using the available commodity hardware.
1 November 2020 in Computer Science and Information Technologies
SSShantharajah Shantharajah
Shantharajah Shantharajah
Iris recognition is a promising biometric authentication approach and it is a very active topic in both research and realistic applications because the pattern of the human iris differs from person to person, even between twins. In this paper, an optimized iris normalization method for the conversion of segmented image into normalized form has been proposed. The existing methods are converting the Cartesian coordinates of the segmented image into polar coordinates. To get more accuracy, the proposed method is using an optimized rubber sheet model which converts the polar coordinates into spherical coordinates followed by localized histogram equalization. The experimental result shows the proposed method scores an encouraging performance with respect to accuracy.