The accuracy of using optimal parameter values in kernel functions is as a determinant to obtain maximum accuracy results on Image retrieval with Support Vector Machine (SVM) classification. Experiments conducted in this study aimed to obtain optimal Gaussian / Radial Basis Function (RBF) kernel function parameter values on non-linear multi class Support Vector Machine (SVM) method. Cross Validation and Grid Search methods were applied in analyzing and testing the optimization range of SVM-RBF kernel parameter values to recognize the image of Indonesian traditional Batik which has geometric decorative patterns. In addition, a feature dataset of Batik images from the results of Discrete Wavelet Transform (DWT) level 3 db2 was used in this study. The feature dataset was used as training and test dataset. By using Cross validation and Grid Search, it resulted in the range value of parameter C = {2 6.5 , 2 6.75 , 2 7 , 2 7.25 , 2 7.5 , 2 7.75 , 2 8 } and γ ={2-14.5 , 2-14.75 , 2-15 , 2-15.25 , 2-15.5 , 2-15.75 , 2-16 }, and the accuracy value of maximum classification for parameter C = 2 7 and γ=2-15. These range results of parameter values and optimal parameter values can be used as a reference in applying parameters on image recognition with geometric decorative motif texture using SVM-RBF kernel classification.

This research proposes an encryption method on images using a combination of chaotic methods, streams, and hash functions. SHA-1 is used as a hash function to encrypt key inputs to be more secure and can produce more dynamic keys at chaotic and stream encryption stages. Chaos encryption is done by dividing the image into small blocks where each encrypted block differs based on a dynamic key pattern based on chaotic keys. At the last stage, all blocks are made as whole images again to be encrypted by the stream method. Tests carried out on standard RGB images and Indonesian batik images. Encryption quality measurements using entropy, histogram analysis, UACI, NPCR, SSIM, PSNR, and the avalanche effect. Based on the results of trials the proposed method is proven to be resistant to various attacks such as statistics as evidenced by the average entropy value of 7.9996, avalanche effect value of 50.0366 and a relatively uniform histogram, while differential attack as evidenced by the value of UACI 33.5716 and NPCR 99.6082 where this value is very close to ideal. Also visually the results of the encryption look very chaotic and very different from the original image, which is evidenced by the value of PSNR 8.0191 and SSIM 0.0081. The decryption process can also be done perfectly wherein the resulting infinity value on PSNR and value 1 on SSIM.

Support Vector Machine (SVM) with Radial Basis Functions (RBF) kernel is one of the methods frequently applied to nonlinear multiclass image classification. To overcome some constraints in the form of a large number of image datasets divided into nonlinear multiclass, there three stages of SVM-RBF classification process carried out i.e. 1) Determining the algorithms of feature extraction and feature value dimensions used, 2) Determining the appropriate kernel and parameter values, and 3) Using correct multiclass method for the training and testing processes. The OaO, OaA, and DAGSVM multi-class methods were tested on a large dataset of batik motif images whose geometric motifs with a variety of patterns and colors in each class and containing similar patterns in the motifs between the classes. DAGSVM has the advantage in classification accuracy value, i.e. 91%, but it takes longer during the training and testing processes.

Classification in recognizing image of letters and numbers is useful to recognize vehicle license plates. This study aims to maximize classification accuracy value of feature extraction method using matrix segmentation. The dataset consists of 300 vehicle license plate images which have 36 classifications, 26 classes for A-Z letters image, and 10 classes for 0-9 numbers image. The research stages carried out to maximize the results of the classification using multiclass SVM-RBF nonlinear are: preparing region image of interest, image enhancement, image segmentation of letters and numbers, determining the best n value for n x n matrix segmentation, calculating total points of each segment as feature value, and determining the best value for C and γ as the value of RBF kernel parameter. The result of this study shows a maximum value of 92% classification accuracy using n = 5, γ = 0.8, and C = 15.

Query optimization is an important task in a client/server environment of a distributed database, whereas a health epidemiologist data distribution based on DBD data on Geographic Information Systems (GIS). A proper method for a particular query process function is needed to generate query optimization on a distributed database. The query process requires important attention especially in distributed databases because the result of a cost-based query process is accessed by involving a number of attributes and visited sites. A query operation typically will search for data from various attributes in a scattered database table, although the processes do not require all table attributes. Query optimization requires minimum query operating costs (communication costs and access fees). The query cost can be optimized by separating attributes that are not required by the query. This can reduce the amount of communication and access time. The attributes should not be divided indiscriminately to obtain the best result of the query process and a vertical fragmentation method can be used to perform such attribute separation. In this research, attributes separation using vertical fragmentation method for a database health table is studied by comparing Bond Energy Algorithm (BEA) and Graphic Based Vertical Partitioning (GBVP) algorithm. The initial result of vertical fragmentation in both algorithms is the determination of types of attributes separated from a number of specific query process. The result of the separation of attributes from each algorithm is compared and evaluated using Partitioned Evaluator (PE) in order to achieve the access cost of several attributes. The results show that GBVP algorithm is more optimal for use in vertical table fragmentation process applied as query operation on distributed DBD database in a health field. The GBVP algorithm has less computational complexity, results a higher partition evaluator value and has lower query execution time than BEA.