Computer Engineering

In this paper we presents a method for handling multiple metrics and different network parameters simultaneously to analyze the loss factor of routing protocols in mobile ad hoc network(MANET) environments. We have used Taguchi’ loss function to determine the best parameters giving maximum throughput, packet delivery ratio (PDR), average delay, DROP and routing over head simultaneously for AODV protocol. In this paper we have consider various different mobile ad hoc network parameters such as Terrain size, No of Nodes, No of source nodes, Packet transmission rate, Node speed, Pause time, Transmission range, Queue size, Antenna height and receiving power on a multiple signal to noise ratio (MNSR), performance and contribution level of parameters have been analyzed by analysis of variance (ANOVA). The analysis of results shows that the parameters which more affecting the AODV performance in mobile ad hoc networks are Queue size, Receiving power, Source node, Packet transmission rate, Antenna height and transmission range.

Sequential pattern mining has been a focused theme in data mining. Sequential pattern mining algorithms provide better performance for short sequences but are inefficient at mining long sequences, since long sequences generate a large number of frequent subsequences. To avoid the limitations of sequential pattern mining algorithms, closed sequential pattern mining algorithms were developed. Closed sequential pattern mining produces less number of patterns and works more efficiently than sequential pattern mining. In this paper, we propose an efficient algorithm CSPgrow to find out closed sequential patterns. To improve the performance, we developed an Extension Checking pruning method. The results indicate that the proposed algorithm CSPgrow outperforms ClaSP.

Distributed file systems (DFS) is one of the important building blocks for cloud computing environment which supports Map Reduce programming pattern where nodes at the same time serve both computing as well as the storage functions. A file is partitioned into variety of chunk units allotted in different nodes in order that the Map Reduce tasks can be carried out in parallel over the nodes. However, in cloud environment, files can be dynamically performs all the operation like creation, deletion, and modification. These consequences in load imbalance on the storage resources; that is, the different file modules are not distributed as consistently as possible among the nodes. A fully distributed load balancing algorithm is offered to manage with the load imbalance problem and thus it increases the overall performance of the system.

The present study deals with the performance and comparison between two architecture in neural networks namely perceptron neural networks and feed forward neural networks. This networks help to identify the particular species of olive ridley sea turtle. The existing algorithm in training the images faces many difficulties such as time delay in training the images and also improper learning rate in the ANN. To overcome this problems the new algorithm is developed were training is made with both perceptron and feed forward neural networks. Initially 8 images were trained and these images consisting of four main features which includes length, breadth, color and shape. These trained images are processed for performance analysis to make effective identification of olive ridley sea turtle.

Mobile Adhoc Networks (MANETs) as the name signifies is a network formed by collection of mobile adhoc devices (nodes). MANET is an autonomous decentralized wireless network where each node is free to move anytime anywhere. Routing always being the most researched topic in the cases of networks. Routing in MANETs is mainly of two types proactive and reactive. This paper presents a route discovery technique used in the reactive routing protocol i.e. AODV. Simulation results using NS2 are also shown which proves that this route discovery technique works better than the original route discovery used earlier.

In cloud storage, the server that stores the client’s data is not trusted. Users would like to check if their data has been tampered or deleted. Provable data possession is used for security by read the file for hash and later specifies its location to verify by applying message digest algorithm which results in generating hash code. Generated hash code is compare with the other hash code generated on different storage. Results showed that precompressed data’s that is the data which is designed by the precompressed technique has their difference in message digest present at local connection side and secure connection side at both secure and insecure server.

In general, existing cQA forums provide only textual answers, which are not informative enough for many questions. Conventional MMQA aims to seek multimedia answers without the assistance of textual answers. This MMQA scheme using diversification method is able to enrich textual answers in cQA with appropriate meta data which improves the informativeness and it bridges two gaps (i) The gap between questions and textual answers (ii) The gap between textual answer and multimedia answer. This scheme consists of three components: answer medium selection, query generation and data selection and presentation. This scheme determines the type of media information that can be added for a textual answer. This scheme predicts the type of medium to be added using Naïve Bayes Classifier, it automatically generates query based on QA knowledge and performs multimedia search. Finally this scheme performs query adaptive re-ranking and duplicate removal to obtain a set of images and videos for presentation along with textual answer. This scheme uses diversification methods to make the enriched media data more diverse. This scheme uses Page Ranking algorithm to retrieve more relevant and diverse search results.