ebook img

Feb 2018 PDF

140 Pages·2017·10.29 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Feb 2018

“Research is Creating New Knowledge” - Neil Armstrong SARJAN A Peer Reviewed Research Journal Published By Silver Oak Group of Institutes VI Edition, 2017-18 Chief Patrons Mr. Shital Agrawal Mr. Janak Khandwala Mrs. Poonam Agrawal Dr. Sweta Khandwala Patrons Dr. Saurin Shah Dr. Siddharthsinh Jadeja Editors Prof. Dharmesh Tank Dr. Rita Sharma Prof. Dipika Gupta Reviewers Dr. R. A. Thakker, Vishwakarma Govt. Engineering College Dr. N. M. Bhatt, Principal, Gandhinagar Institute of Technology Captain Dr. C. S. Sanghavi, L. D. College of Engineering Dr. Rita Sharma, Silver Oak College of Engineering & Technology Dr. Bhavik Suthar, L. D. College of Engineering Dr. Amit Agrawal, Aditya Silver Oak Institute of Technology Dr. Pina Bhatt, Silver Oak College of Engineering & Technology The Journal SARJAN is a peer reviewed research journal yearly published by Silver Oak Group of Institutes. The present issue is the sixth edition of SARJAN. It publishes thoughtful contributions from researchers that offers insight and perspective, knowledge and understanding of multi-disciplinary research in areas of various disciplines of Engineering ; Computer Engineering, Information and technology, Electronics and Communications, Mechanical Engineering, Aeronautical Engineering, Civil Engineering, Chemical Engineering, Electrical Engineering and also Applied Sciences and Humanities. The mission of the journal is to foster research culture and increase research productivity of the faculty members in the institute by promoting and publishing their research skills. The institute believes that research in academics can only solve the problem of employability in the country and cater to the needs of the industry. SARJAN has consistently since 2012 provided a platform to researchers and academicians to contribute in this pool of knowledge that helps the aspiring researchers and students to update themselves with latest innovations and knowledge and thus enhance their technical skills. FORWARD Dear friends, I am happy to note that as a regular feature Silver Oak Group of Institutes is publishing it’s yearly research journal SARJAN. Being a young institution and doing such motivating research activities, it's really excellent and indeed a proactive step as it is the need of a day and especially for an academic institute which is aspiring to be a university. We know that 21st century is the era of knowledge application and such initiatives will empower the faculty members to exhibit their research skills which will also benefit the student community in present competitive times. Former Vice Chancellor of Gujarat University Ahmedabad. INDEX Sr. Paper Title Authors Page No. No 1. "A New Approach For Improving Accuracy Of Multi Prof Aniket Patel, 1 Label Stream Data Using Classification Prof. Jaimin Dave, And Clustering" 2. Service improvement in Nodal Failover using Luci (Red Prof. Rahul Shrimali 6 Hat Cluster System)” 3. A Survey on Facial Expression Recognition Techniques Vandana Patidar 8 Prof. Dharmesh Tank 4. Real Time Application Based Approach for Customizable Dr. Dushyantsinh B. 13 Log Mining Rathod Mr. Sagar Patel Mr. Manish Singh 5. Emerging Trends of Clustering in Web Usage Prof. Archana 18 Mining and It’s Business Applications Gondalia Dr. Dushyantsinh Rathod Prof. Namita Patel 6. Rotation Perturbation Technique for Privacy Preserving Prof.Aniket Patel, 23 in Data Stream Mining Prof. Dharmesh Tank, Kalyani Kathwadia 7. Analysis and Design of WR75 Coaxial to Rectangular Prof. Krupa Prajapati 27 Waveguide Adapter for KU Band 8. Techniques in Driver Fatigue Recognition Using Prof. Tulsi A. Pandya 29 Image Processing: A Review 9. Implementation of wavelength diversity technique on Prof. Rina Parikh 33 free space optical communication link for mitigating the Rachana Jadvani atmospheric turbulence. A Review 10. Survey Of Speaker Identification System Using Jui Trivedi , 38 Watermarking Technology For Spoofing Prof. Nikunj V. Attack Of Voice Conversion Tahilramani 11. Survey on Information hiding in speech signal Ruchi Patel, 42 Prof. Nikunj V. Tahilramani 12. Survey on Bandwidth Extension Techniques of Janki Patel, 46 Speech Signals Prof. Nikunj V. Tahilramani 13. PVD and HVOF coatings for industrial Dr Pina Bhatt, 51 applications-A Review Parthiv Trivedi 14. Vapour absorption refrigeration system using waste Prof. Kaushik 56 heat recovery for engine cooling system Savaliya 15. Design, development and analysis of impeller for Porf. Kataria 60 sand slurry pumping Mahendra B., Prof. Rushil Shah 16. Finite Element Analysis and Experimentation of Prof. Kataria 63 Carbon Fiber Chain Drive Mahendra B. 17. An Article on Ancient Indian Mechanical Engineering Dr Pina Bhatt, Prof. 68 “Yantrashastra” Meet Shah INDEX 18. A New Approach: Teaching Second Prof. Nikita Gadani 70 Language through Poetry 19. The Evil of Manipulations in the Gold Market Prof. Dhwani Shah 77 20. Study of Aerosol Optical Depth at Udaipur Dr. Roshni Dave, 81 Prof. Harsh Solanki 21. Pedagogy On Engineering Mathematics And Pure Mr. Jishan K Shaikh, 83 Mathematics Education Mr. Dhruv A Pandey 22. Reliability Analysis Of A Stand-By System With Dr Shailja Sharma 89 Environment Failure 23. An Inequality Based Proof Of The Fixed Point Theorem Dr. Harsh Trivedi 95 In C*-Algebra Valued Modular Spaces 24. Untapped Potential of Multilevel Car Parking Prof. Aakash B 98 Desai 25. Rooftop rainwater harvesting: A case study at Sagar A. Bhatia, Prof. 101 Silver Oak Campus, Ahmedabad Mrunalini H. Rana 26. Effect of Pollution on City Environment: A Porf. Akash 106 Discussion on Gujarat Scenario Bhardwaj 27. Comparison of Structural Systems for Prof. Parth Danani, 112 Composite Construction in High Rise Building Prof, Hasumati Patel 28. Traffic Management in Urban Area Lect. Anuj Bhatt, 117 Lect. Reecha Panchal 29. A Review on Desalination of Sea Water Vidhi Bhavsar, 120 Nikhil Tolwani 30. Rehabilation Of Unlined Canal To Lined Prashant Chavda, 122 Canal A Case Study @ “Bavala Canal Chirag Rathod, Ankit Makwana, Kalp Patel, Prof. Mrunalini H. Rana 31. Determining the possibility of reusing exhaust air from Prof. Niharika 125 spray dryer for energy Mehta conservation 32. Experimental Study for Optimal Power Flow of Prof. Manan Y. 129 IEEE 14 Bus System using Static VAR Pathak Compensator Device with Voltage Constraints ISSN: 2320-2122 Edition: VI 2017-18 A NEW APPROACH FOR IMPROVING ACCURACY OF MULTI LABEL STREAM DATA USING CLASSIFICATION AND CLUSTERING 1 2 Prof Aniket Patel , Prof. Jaimin Dave , 1,2 Assistant Professor, Information Technology Department, Silver Oak College of Engineering and Technology, Ahmedabad, Gujarat Abstract: Data Stream Mining is the process of 1.3 Stages in Data Analysis. extracting knowledge structures from continuous, rapid data records. The recent advances in The intelligent data analysis has passed through a hardware and software have enabled the capture of number of stages. Each stage addresses new kind different measurements of data in a wide range of of research issues that have arisen. fields. These measurements are generated continuously and in a very high fluctuating data a) Statistical exploratory data analysis: rates. It includes in sensor networks, web logs, and The goal was to explore the available data in computer network traffic. The storage, querying order to test a specific hypothesis. With the and mining of such data sets are highly advances in computing power, machine computationally challenging tasks. learning field has arisen. 1.0 Introduction b) Machine Learning: The objective was to find computationally In recent years, advances in hardware technology efficient solutions to data analysis problems. have facilitated the ability to collect data Along with the progress in machine learning continuously. Simple transactions of everyday life research, new data analysis problems have such as using a credit card, phone or browsing the been addressed. Due to the increase in web lead to automated data storage. Similarly, database sizes, new algorithms have been advances in information technology have lead to proposed to deal with the scalability issue. large flows of data across IP networks. When the Moreover machine learning and statistical volume of the underlying data is very large, it leads analysis techniques have been adopted and to a number of computational and mining modified in order to address the problem of challenges: very large databases. 1.1 Data Stream c) Data Mining: Uninterrupted flow of a long sequence of data, Data mining is use of automated data analysis such as in audio and video data files. Simple techniques to uncover previously undetected transactions of everyday life such as using a credit relationships among data items. Data mining is card, a phone or browsing the web lead to that interdisciplinary of (or between more than automated data storage. Similarly, advances in one branch of learning) field of study that can information technology have lead to large flows of extract models and patterns from large data across IP networks are data streams. amounts of information stored in data repositories. 1.2 Database Vs Data Stream In Databases or other conventional programs, data d) Data Stream Mining: is resident and input rate is under the control of the In this model, data does not take the form of programmer. In newer applications data input as persistent relations, but rather arrives in continuous, ordered data streams. multiple, continuous, rapid, time-varying data Examples: Network monitoring and traffic streams. Data Stream Mining is the process of engineering, Telecom call records, financial extracting knowledge structures from applications, Sensor networks, Web logs and click continuous, rapid data records. streams. Data elements in the stream arrive online, system has no control over arrival order, either 2.0 Data Mining within a data stream or across many streams. Data The dictionary meaning of mining is “excavation” streams are potentially bounded in size, once an or “abundant source (of information etc.)”. It’s element from a data stream has been processed; it about a decade or two years ago advances in is discarded unless otherwise archived. hardware and software have enabled the capture of different measurements of data in a wide range of SILVER OAK GROUP OF INSTITUTES 1 ISSN: 2320-2122 Edition: VI 2017-18 fields. These measurements are generated continuously and in a very high fluctuating data rates. Examples include sensor networks, web logs, and computer network traffic. 2.1 What Is Data Mining? Data mining (sometimes called data or knowledge discovery) is the process of analyzing data from different perspectives and summarizing it into useful information - information that can be used to increase revenue, cuts costs, or both. Technically, data mining is the process of finding correlations Figure 3.1: Mining Data Stream Process or patterns among dozens of fields in large relational databases. 1.1 DATA BASED TECHNIQUES 2.2 Applications: Data-based techniques refer to summarizing the Data mining is primarily used today by whole dataset or choosing a subset of the incoming companies with a strong consumer focus - stream to be analyzed. Sampling, load shedding retail, financial, communication, and and sketching techniques represent the former one. marketing organizations. It enables these companies to determine relationships among Synopsis data structures and aggregation represent "internal" factors such as price, product the later one. positioning, or staff skills, and "external" factors such as economic indicators, 1.2 TASK-BASED TECHNIQUES: competition, and customer demographics. And, it enables them to determine the impact Task-based techniques are those methods that on sales, customer satisfaction, and corporate modify existing techniques or invent new ones in profits. Finally, it enables them to "drill order to address the computational challenges of down" into summary information to view data stream processing. Approximation algorithms detail transactional data. With data mining, a and sliding window represent this category. retailer could use point-of-sale records of customer purchases to send targeted promotions based on an individual's purchase I.APPROXIMATION ALGORITHM: history. By mining demographic data from comment or warranty cards, the retailer could These algorithms can result in an approximate develop products and promotions to appeal to solution with error bounds. The idea is that mining specific customer segments. For example, algorithms are considered hard computational Blockbuster Entertainment mines its video problems given its features of continuality and rental history database to recommend rentals speed and the generating environment that is to individual customers. American Express featured by being resource constrained. can suggest products to its cardholders based Approximation algorithms have attracted on analysis of their monthly expenditures. researchers as a direct solution to data stream mining problems. However, the problem of data 1. THEORETICAL FOUNDATIONS OF rates with regard with the available resources could STREAM MINING not be solved using approximation algorithms. Research problems and challenges that have been arisen in mining data streams have its solutions II.SLIDING WINDOW: using well established statistical and computational Many of the synopsis structures discussed use the approaches. We can categorize these solutions to entire data stream in order to construct the data-based and task-based ones. In data-based corresponding synopsis structure. The inspiration solutions, the idea is to examine only a subset of behind sliding window is that the user is more the whole dataset or to transform the data vertically concerned with the analysis of most recent data or horizontally to an approximate smaller size data streams. The sliding-window model of computation representation. At the other hand, in task-based is motivated by the assumption that it is more solutions, techniques from computational theory important to use recent data in data stream have been adopted to achieve time and space computation. Thus the detailed analysis is done efficient solutions. Figure 3.1 shows the general over the most recent data items and summarized. processing of mining data streams. versions of the old ones. Therefore, the processing SILVER OAK GROUP OF INSTITUTES 2 ISSN: 2320-2122 Edition: VI 2017-18 and analysis is only done on a fixed history of the mean is the most frequent used exclusive clustering data stream. algorithm. The procedure of k-means follows a simple way. Idea is to determine K centers, one for 4. MINING TECHNIQUES each centers. The main challenge in this algorithm Mining data streams has attracted the attention of is to place these centers in a clever way. data mining community for the last seven years. A 1. Choose the number of k clusters number of algorithms have been proposed for 2. Randomly assign items to the k clusters extracting knowledge from streaming information. 3. Calculate new centroid for each of the k clusters In this section, we review clustering and 4. Calculate the distance of all items to the k classification. centroids 5. Assign items to closest centroid 4.1CLUSTERING: 6. Repeat until clusters assignments are stable In Today’s world, rapid growth of computer technology also increase the use of data in a distributed environment. To match up this speed of a clustering of data stream becomes a natural process. Clustering is a process of organization objects into groups whose members are similar in some way. In other words” Clustering is a useful and ubiquitous tool in data analysis, is the problem of finding a partition of a data set so that under some definition of similarity similar items are in the same part of the partition and different items are in different parts” The main aim of clustering on streaming data is to partitioning a list of data points into k groups of “similar” objects by The K-median streaming algorithm scanning the data once. The main challenge in clustering is to determine an Firstly It solves k-median on Xi using LSEARCH intrinsic group in a set of unlabelled data. But there ,Secondly X’ centers obtained from chunks one is no perfect criterion which decides the final aim through I iterations of the steam where each center of clustering. It is the user that decides this c obtained by clustering Xi is weighted by the criterion which supplies the result of clustering that number of points in Xi assigned to c. Lastly output suit their need. The problem of clustering uncertain the k centers obtained by clustering X’ using data streams is especially challenging because the LSEARCH . Experiments shows that the STREAM uncertainty in attribute values can significantly produces a solution whose cost is almost constant affect the clustering behavior of the data points. times the cost would get by applying L search Our motivation here is to study different directly to stream. procedures or algorithms through which we can 4.2.CLASSIFICATION: cluster data efficiently. Clustering of data stream is not possible every time because the data records The problem of classification is perhaps one of the generated are noisy or incomplete due to unreliable most widely studied in the context of data stream distributed system. mining. The problem of classification is made more difficult by the evolution of the underlying data Clustering Algorithms stream. Therefore, effective algorithms need to be K-Means Clustering Described by Guha, Mishra, designed in order to take temporal locality into Motwani and O'Callaghan achieve a constant factor account. Data stream classification algorithms approximation for the k-Median problem in a require appropriate and complete evaluation single pass and using small space. It requires O(nk) practices. The evaluation should allow users to be time and O(n_) space where “k” is the number of sure that particular problems can be handled, to centers, “n” is the number of points and _<1. They quantify improvements to algorithms, and to have proved that any k-median algorithm that determine which algorithms are most suitable for achieves a constant factor approximation cannot their problem. The goal of classification is to achieve a better run time than O(nk). The algorithm produce a model that can predict the class of starts byclustering a calculated size sample unlabeled examples, by training on examples according to the available memory into 2k, and whose label, or class, is supplied. To clarify the then at a second level, the algorithm clusters the problem setting being addressed, several above points for a number of samples into 2k and assumptions are made about the typical learning this process is repeated to a number of levels, and scenario: finally it clusters the 2kclusters into k clusters. K- SILVER OAK GROUP OF INSTITUTES 3

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.