Roheet Bhatnagar

Implementation of Particle Filters for Single Target Tracking Using CUDA

In order to implement Sequential Bayesian estimator using Monte carlo simulation and to get rid of limitations of Kalman filter, Particle filtering techniques plays a very crucial role for target tracking applications in state space where Importance sampling approximately distributed by posterior distribution with multimodel feature and robustness to noise. However as the particles becomes very large, the Monte Carlo representation becomes nearly equivalent to analytical description characterization for posterior distributions and has some deficiencies such as high computational cost and low sampling efficiency. Therefore, emerging computing platform, CUDA may be regarded as most appealing platform for such implementaion. Representation provided with set of samples for target distribution of state leads to increase in sampling efficiency so that graphics processing unit (GPUPU) becomes more appealing to use in Particle filter. The modification on mapping architecture are evaluated with qualitative analysis. The proposed design will be 3.5 times faster than direct design.

Distributed Medical Image Management: A platform for storing, analysis and processing of Image Database over the cloud

In recent years Medical Informatics related apps, combine information technology with raw and clinical Medical data to improve the standards in health care diagnosis and Medical Research. This paper mainly deals with how the environment for storing, analysis and processing of Image Databases will be created and operated over the networks. In addition to that different considerable elements like network protocols, Grid based distributed computing systems and accessible elements are explained in this paper. Cloud computing brings radical changes in Medical Image Database management systems operating over the wide area of networks across the Globe. In a holistic way this frame work enables researchers and doctors to design and operate Image Processing Application in a simple, economy manner. It also enables them to make best use of Cloud and other network resources as they are available at real time situations.

Surface Mining Signal Discrimination Using Landsat TM Sensor: An Empirical Approach

In Chotanagpur plateau of Jharkhand State in India, mining is a prominent activity. Sample sites of three such ores, viz. Bauxite, Hematite and Uranium were taken up for the present study wherein the first two are extracted through surface mining, leaving their signatures on the earth’s surface, while the third one, extracted through underground mining process, leaves its trail in tailing-pond, after its beneficiation, because of the higher degree of its radioactive property (half-life of Uranium is around 4,500 million years [1]). The Study attempts to statistically discriminate mining signals that were picked up as DN (digital number) values of the first four spectral bands of TM (Thematic Mapper) sensor, displayed on the graphic screen as the additive colour composite, using three primary colours namely red, green and blue (RGB) as standard FCC (false colour composite) of Landsat satellite-image. The said discrimination were based on application of two independent statistical algorithm on these values, one being paired-sample Student’s t-Test (at 95% confidence level and 3°of freedom) through SPSS (ver. 19) software and the second, subsequently, ANOVA (Analysis of Variance) test (at 95% confidence level), in order to further discriminate the signals based on parameters like spectral bands and nature of mineral being mined. According to the first algorithm Bauxite was found to be clearly discriminated, both from Uranium as well as Hematite, while Hematite could only be distinguished from Bauxite but not from Uranium. Performance of ANOVA test on the DN values discriminated the surface mining signals pertaining to these three different ores and it showed a high variance between the spectral bands both within the same ore group emphasising that different bands of the satellite sensors specifically identify features and also between the different ore groups.

Analysis of Online News Popularity and Bank Marketing Using ARSkNN

Data mining is a process of evaluating practice of examining large preexisting databases in order to generate new information. The amount of data has been growing at an enormous rate ever since the development of computers and information technology. Many methods and algorithms have been developed in the last half-century to evaluate data and extract useful information to help develop faster. Due to the wide variety of algorithms and different approaches to evaluate data, several algorithms are compared. The performance of any algorithm on a particular dataset cannot be predicted without evaluating it with the same constraints and parameters. The following paper is a comparison between the trivial kNN algorithm and the newly proposed ARSkNN algorithm on classifying two datasets and subsequently evaluating their performance on average accuracy percentage and average runtime parameters.

An Approach for Adapting Component-Based Software Engineering

Traditionally, software was developed by writing a main method which invoked many subroutines. Each subroutine was programmed as a specific part of the program based on the given requirements and function partitions. Software engineers called for enhanced software quality, timely, at reduced costs and hence adopted the use of reusable components. This work intends at designing and augmenting generic software components for admission management system domain using OOPs methods. The analysis of major admission management system functions, data and behaviors has been taken herewith. Also, pattern-based domain engineering was conducted so as to identify the structure points thereby factoring out generically reusable components.

Machine Learning and Big Data Processing: A Technological Perspective and Review

This paper discusses the role of Machine Learning (ML) based algorithms and methods in Big Data Processing & Analytics (BDA). ML and BDA are both evolutionary fields of computing and the developments in these fields are complementing each other. The ever changing data landscape in modern digital world have resulted in newer ways of data processing frameworks in order to get meaningful insights which are unprecedented. This paper presents a detailed review on latest developments in ML algorithms for Big Data Processing. In later section key challenges associated with application of ML based approaches are also discussed. ML based Big Data Processing has gained popularity and new developments are on the rise for efficient data processing. This field is witnessing unparalleled emergence of new methods and approaches for efficient data processing in order to discover interestingness for decision making. Thus, more and more ML based data processing approaches are being used for Big Data Processing. With the splurge data from different newer sources, heterogeneous nature of data, uncertain & unstructured data, the so called Big Data with all its characteristics (5 Vs) there is an ever increasing need to use approaches which aid in modelling and processing of these data, provide automated approach to data processing and so on. These type of new processing requirements have given a big boost to the development of new ML based methods for managing & processing them. The paper will be useful to the scholars who are researching in this interesting & challenging domain of ML and Big Data Processing.

Analysis of Credit Risk Prediction Using ARSkNN

Credit risk is characterized as the risk that borrowers will neglect to pay its advance commitments and loan obligations. It is very hard to predict the outcomes (risky borrower) manually as the evaluation of large features set is quite time consuming. That’s why, we need some good predictor as classifier. The traditional k-NN is one pre-established classifier used in various domains along with credit risk predictions. The newly conceptualized ARSkNN is another such classification which reduces the runtime in predicting the outcomes and improves overall accuracy percentage of the predicted classes over Traditional k-NN. The method adopt the similarity measure which is based on the Mass estimation rather than distance estimation for predicting the K- nearest neighbor. The results were compared using WEKA 3.7.10 as tool and found significant improvement vis-a-vis the evaluation parameters by the ARSkNN method.

A New Method for Minimizing Unnecessary Handoff in 802.11

In a mobile wireless network, the unnecessary handoff is the main issue that effects the performance of the network. A handoff occurs when the Received Signal Strength (RSS) is less than the threshold RSS and the new RSS (from neighbors Access point) is more than the present RSS. Earlier the handoff decision algorithms consider RSS only which results in an increase in handoff counting or handoff rate but in a wireless communication system, the signal strength varies due to scattering, shadowing and reflection because of obstacles in the network region. This variation in RSS leads to unnecessary handoff. In this paper, a new method for handoff decision is proposed to reduce the unnecessary handoff. With simulation results or mathematical analysis 25% unnecessary handoff reduced.

Can We Use Mass-Based Similarity Measure in Classification?

Similarity measures are very much essential in solving many data mining tasks such as clustering, information retrieval, and classification. A large number of the similarity measures directly or indirectly depend upon distance. Recently developed mass-based similarity measure, Massim, is well established in information retrieval task with algorithm MassIR. This paper will examine the probable uses of mass-based similarity measure in classification tasks.