http://localhost:8080/xmlui/handle/123456789/2154 2026-06-23T06:19:03Z http://localhost:8080/xmlui/handle/123456789/3439 Title: Efficient Embedding in B&W Picture Images Authors: Chhajed, Gyankamal J.; Shinde, S.A. Abstract: This paper proposes novel method of embedding and extraction of data in Black and White picture images. The main focus of this method is on steganography in Black & white picture image. This method embeds more number of bits in a block as compared to earlier methods, which are limited to one or two bits. This method suggests the searching of the block of suitable size and choosing it for data embedding. The selection of block size is dependent on the encrypted secret message pattern which is encrypted using a secret key shared by sender and receiver and the pattern of block. Here main aim is to utilize the image as much as possible with its own pattern of black and white pixel. In this method very first the encrypted message pattern is taken and it is matched with the block size of 2X2, 3X3 and like wise. The blocks which are giving maximum matching with the bits of the encrypted message is selected for the embedding. To maintain the visual quality of image we will only change at the maximum 2 pixels in the block if size is more than 2X2 and 1 pixel for 2X2 block. To extract the correct data we will use the odd even feature of the block of size 3X3, which will be utilized to keep the embedding information that is size of the message , block size and locations where data is embedded. This information is also encrypted with secret key for security purpose and embedding will start from the end of the image that is right bottom corner and scanning and embedding will be from right to left and bottom to top. This information is used to extract exact amount of data without the need of checking total image. The secret message is extracted after decoding extracted bits from the stego-image with secret key without using original image accurately. This method will prove to be good for embedding increased capacity of data without causing much distortion 2010-04-25T00:00:00Z http://localhost:8080/xmlui/handle/123456789/3435 Title: Defect classification as problem classification for Quality control in the software project management by DTL Authors: Hanchat, D.B.; Sayyad, Shabina; Shinde, S.A. Abstract: There are various reasous and causes which lead to failure of software that may come right from it's starting point of requirement analysis up to launching of product in the market .One has to do the root cause analysis of software failure so that these failures should not be reproducible. There are various problems due to which the software may give the bugs, errors, fault and ultimately the failure. Enlisting the problem, analyzing the problem after reporting is must before the fixing of the problem and going into the root cause of the problem. The classification of the problem will definitely help us to sort out the problems and will help to go to the root of problem. Once problems have been reported it can be classified by using any classification method depending upon the properties and their values. We do combine the decision tree learning with the input as the current problems. Decision Tree will be trained with trainee example with similar type of problems, their "properties and values". The DTL will help us to classify the problem and ultimately give us the sorted problems to do analysis of problem [2]. Analysis and classification of the problems will also help in the quality control of the product. We have taken defects classification as an example in this paper. 2010-10-23T00:00:00Z http://localhost:8080/xmlui/handle/123456789/2797 Title: Use of Instance Typicality for Efficient Detection of Outliers with Neural Network Classifiers Authors: Sane, Shirish S.; Ghotal, Ashok S. Abstract: Detection of outliers is one of the data pre-processing tasks. In all the applications, outliers need to be detected to enhance the accuracy of the classifiers. Several different techniques, such as statistical, distance-based and deviation-based outlier detection exist to detect outliers. Many of these techniques use filter method. A wrapper method using the concept of instance typicality may also be used to detect outliers. This paper deals with a new wrapper method that builds an initial model using neural networks and treats values at the output of neurons in the output layer as the typicality scores. Instances with lowest output values are treated as potential outliers. In addition, the method is also useful to build compact and accurate classifiers by selecting a few most typical instances resulting in significant reduction in storage space. The method is generic and thus can also be used for instance selection with any kind of classifiers. Resultant compact models are useful for imputation of missing values. 2006-07-01T00:00:00Z http://localhost:8080/xmlui/handle/123456789/2155 Title: A novel Supervised Instance Selection algorithm Authors: Sane, Shirish S; Ghatol, Ashok Abstract: Instance selection is often used in case of lazy classifiers. This paper addresses the need of instance selection in case of neural network and decision tree classifiers and presents a novel Supervised Instance Selection (SIS) algorithm. Initially, a neural network classifier is constructed using all training instances. The algorithm then selects a few instances using the certainty values of the wrapped neural network to construct a compact classifier. Empirical study made with standard datasets shows that SIS save on 70% of storage space without degrading the accuracy. It is independent of nature of the dataset and the tool used. 2007-01-11T00:00:00Z