Informatics Publishing Limited

K. Kiran ¹, Nazima Begum ¹, R. Ramya ¹, Abhishek Alfred Singh ¹, P. Deepa Shenoy ¹, K. R. Venugopal ², Lalit M. Patnaik ³

Affiliations
1 Department of CSE, University Visvesvaraya College of Engineering, Bangalore, IN
2 University Visvesvaraya College of Engineering, Bangalore, IN
3 Indian Institute of Science, Bangalore, IN

Abstract

WiFi and WiMAX are two popular wireless standards, where WiFi is a wireless LAN standard and WiMAX is a wireless MAN standard. WiFi (IEEE 802.11) provides coverage of few hundred feet whereas WiMAX (IEEE 802.16) covers a range of upto 40 miles. Nowadays multiple radios are equipped within a single device. One way to achieve high performance in a network consisting of such multi-radio devices is by splitting traffic over the multiple radios. Research work has shown that traffic splitting has a positive impact on network throughput. In this work, our objective is to split the data traffic dynamically in an ad-hoc network with such hybrid nodes equipped with WiFi and WiMAX radios. We have used Bee-Hive Routing Protocol to make routing decisions in the network and analyze the effect of traffic splitting on network throughput.

Keywords

Bee-Hive Routing Protocol, Dynamic Traffic Splitting, Hybrid Node, LAN, MAN, WiFi, WiMAX.

Full Text

     

S. Bharathi ¹, P. Deepa Shenoy ², K. R. Venugopal ³, L. M. Patnaik ⁴

Affiliations
1 Department of MCA, Dr Ambedkar Institute of Technology, Bangalore, IN
2 Department of Computer Science and Engineering, University Visvesvaraya College of Engineering, Bangalore, IN
3 University Visvesvaraya College of Engineering, Bangalore University, Bangalore, IN
4 Indian Institute of Science, Bangalore, IN

Abstract

The task of indentifying similar objects within the querying image remains challenging. It is due to viewpoint and lighting changes, deformation and partial occlusions that may exist across different examples. In this framework we focus on combination methods that ensemble multiple descriptors at multiple spatial resolution levels. Ensemble PHOG (Pyramid histogram orientation and gradient) and SIFT (Scale invariant feature transformation) descriptors are used for the feature extraction to achieve the good classification accuracy. Within a region local feature was captured by the distribution over edge orientation, and spatial layout by tiling the image into regions at multiple resolutions. The SIFT features are extracted for each PHOG block. These features are trained using SOM network. Later SVM and Neural network classifiers are used for classification. Results demonstrating the effectiveness of the proposed technique are provided using confusion matrix, transition matrix and other accuracy measures. Area of different land cover regions are calculated, which can be used for land use changes.

Keywords

PHOG, SIFT, Classification, Satellite Image.

Full Text

     

J. S. Saleema ¹, P. Deepa Shenoy ², K. R. Venugopal ³, L. M. Patnaik ⁴

Affiliations
1 Dept. of Comp. Sci., Christ University, Bangalore, IN
2 Department of CSE, University Visvesvaraya College of Engineering, Bangalore, IN
3 University Visvesvaraya College of Engineering, Bangalore, IN
4 Indian Institute of Science, Bangalore, IN

Abstract

Cancer prognosis prediction improves the quality of treatment and increases the survivability of the patients. Disease prognosis is identified at the treatment stage and at the recurrence stage. Conventional cancer prediction method deals only with the survival or mortality of the patients, but not with other labels such as severity of the disease through metastasis or multi-primary, stage, grade, etc. The SEER Public Use cancer database has more prominent variables that support better prediction approach. The objective of this paper is twofold. One is to build a prediction model to find the prominent variables by using the standard classifiers and the second is to improve the prediction accuracy through various sampling techniques. The proposed prediction model consist of three phases namely, basic level pre-processing, problem specific processing and modeling classifiers. Problem specific processing phase deals with feature extraction, sampling and response variable selection. The well known classification algorithms (Decision Tree, Naive Bayes and KNN) have been used to model the classifiers for prediction analysis. Apart from the available incident data from SEER (Breast, Colorectal and Respiratory Cancer data) a new mixed combination of the three in equal proportion have been generated for the experimentation. Feature selection through correlation and information gain reduced the attributes to 37 from the raw size of 118. Patient survival, age at diagnosis, stage and multiple primaries in the given order has been identified as the prominent response variable, where as grade performed very low in the experimentation. The performances of various sampling techniques have been studied with the data set size ranging from 500 to 30000 samples for the four prominent labels identified in the previous step. The result shows that the balanced stratified sampling technique always maintains consistency in the performance. Also classifier model with decision tree algorithm optimizes the performance compared to the other algorithms. All the results of the models are tabulated in this paper.

Keywords

Classifier, Pre-Processing, Prognosis Prediction, SEER.

Full Text

     

Chetana Hegde ¹, P. Deepa Shenoy ², K. R. Venugopal ², L. M. Patnaik ³

Affiliations
1 RNS Institute of Technology, Bangalore, IN
2 University Visveswaraya College of Engineering, Bangalore University, Bangalore, IN
3 DIAT, Pune, IN

Abstract

Automated security is one of the major concerns of modern times. Secure and reliable authentication systems are in great demand. A biometric trait like Finger Knuckle Print (FKP) of a person is unique and secure. In this paper, we propose a human authentication system based on FKP image of a person. Depending on the security level required by an organization that implements the proposed system, we provide two modes of security viz. basic mode and advanced mode. In the basic mode, the Radon Transform is applied on pre-processed FKP image and Eigen values are computed. Then we compute the correlation coefficient between the set of Eigen values stored in the database and that of input image to authenticate a person. For advanced level of security, we apply Gabor Wavelet on pre-processed FKP image. The magnitude values of the Gabor Wavelet Transform are computed. Then the correlation coefficient between the set of magnitude values stored in the database and that of input image is used to authenticate a person. For real time implementation, suitable GUI can be developed. The basic mode of security system is found to have FAR as 6.79% and FRR as 0.0517%. The advanced mode has the FAR of about 3.07% and FRR as 1.14%.

Keywords

Correlation Coefficient, Eigen Values, FAR, FRR, Gabor Wavelet Transform, Magnitude Values, Radon Transform, ROC Curve.

Full Text

     

C. N. Pushpa ¹, Gerard Deepak ¹, Mohammed Zakir ¹, J. Thriveni ¹, K. R. Venugopal ¹

Affiliations
1 Department of Computer Science and Engineering, University Visvesvaraya College of Engineering, IN

Abstract

Clustering of similar data items is an important technique in mining useful patterns. To enhance the performance of Clustering, training or learning is an important task. A constraint learning semi-supervised methodology is proposed which incorporates SVM and Normalized Point wise Mutual Information Computation Strategy to increase the relevance as well as the performance efficiency of clustering. The SVM Classifier is of Hard Margin Type to roughly classify the initial set. A recursive re-clustering approach is proposed for achieving higher degree of relevance in the final clustered set by incorporating ENNPI algorithm. An overall enriched F-Measure value of 94.09% is achieved as compared to existing algorithms.

Keywords

Clustering, Constraint Learning, Normalized Pointwise Mutual Information, Recursive Re-Clustering, SVM.

Full Text

     

N. P. Nethravathi ¹, Vaibhav J. Desai ², P. Deepa Shenoy ², M. Indiramma ³, K. R. Venugopal ²

Affiliations
1 Visvesvaraya Technological University, Belagavi-590 018, IN
2 University Visvesvaraya College of Engineering, Bangalore University, Bangalore, IN
3 BMS College of Engineering, Bangalore-560019, IN

Abstract

With the onset of the digital revolution, organizations are increasingly maintaining a huge amount of information on their databases and use data mining tools to extract useful information for their business intelligence. The problem with the availability of the digital information is the scarce privacy leakage. In many business domains, leakage of personal information of the client either directly or through data mining tools can lead to loss of competitive edge of the company, loss of revenue and customer churn. Companies are pushing for encryption and other data transformation methods to keep the data private. But mining tools which invoke algorithms like clustering, classification etc. may not work properly on the transformed data. In this paper, we analyze the privacy preserving data mining solutions and privacy leakage in them through indirect means. The main objective of this paper is to identify the open areas of research on privacy-preserving data mining.

Keywords

Transformation Strategy, Privacy Preserving Data Mining, Cryptography, Wavelet Transformation, Correlation Analysis.

Full Text

     

N. P. Nethravathi ¹, Prasanth G. Rao ¹, Chaitra C. Vaidya ², S. Geethanjali ², P. Madhura ², K. Neha Nandan ², P. Deepa Shenoy ², M. Indiramma ³, K. R. Venugopal ²

Affiliations
1 Visvesvaraya Technological University, Belagavi - 590 018, Karnataka, IN
2 University Visvesvaraya College of Engineering, Bangalore University, Bangalore - 560001, Karnataka, IN
3 BMS College of Engineering, Bangalore - 560019, Karnataka, IN

Abstract

Objectives: Preservation of privacy is a significant aspect of data mining. The main objective of PPDM is to hide or provide privacy to certain sensitive information so that they can be protected from unauthorized parties or intruders. Methods/ Statistical Analysis: Though privacy is achieved by hiding the sensitive or private data, it will affect the data mining algorithms in knowledge extraction, so an effective method or strategy is required to provide privacy to the data and simultaneously protecting the quality of data mining algorithms. Instead of removing or encrypting sensitive or private data, we make use of data transformation strategies that keep the statistical, semantic and heuristic nature of data while protecting the sensitive or private data. Findings: In this paper we studied the technical feasibility of realizing Privacy Preserving Data Mining. In the proposed work, Correlation Based Transformation Strategy for Privacy Preserving Data Mining is used for ordinal data. We apply the method on few datasets namely soybean, Breast Cancer, Nursery dataset and Car dataset. We tabulate the end results applying the proposed strategy on both the original and the transformed dataset and observe correlation difference, Information Entropy and Classification Accuracy with different machine learning algorithms and Clustering Quality. Application/Improvements: As an improvement, the proposed work can be extended by use of vector marking techniques where these techniques help in increasing the efficiency by avoiding unauthorised access to the information.

Keywords

Correlation Analysis, Nominal Data, Ordinal Data, Privacy Preserving Data Mining, Transformation Strategy.

Full Text

   

G. U. Vasanthakumar ¹, R. Priyanka ¹, K. C. Vanitha Raj ¹, S. Bhavani ¹, B. R. Asha Rani ¹, P. Deepa Shenoy ¹, K. R. Venugopal ¹

Affiliations
1 Department of Computer Science and Engineering, University Visvesvaraya College of Engineering, Bangalore University, IN

Abstract

Users of Online Social Network (OSN) communicate with each other, exchange information and spread rapidly influencing others in the network for taking various decisions. Blog sites allow their users to create and publish thoughts on various topics of their interest in the form of blogs/blog documents, catching the attention and letting readers to perform various activities on them. Based on the content of the blog documents posted by the user, they become popular. In this work, a novel method to profile Top Most Influential Blogger (TMIB) is proposed based on content analysis. Content of blog documents of bloggers under consideration in the blog network are compared and analyzed. Term Frequency and Inverse Document Frequency (TF-IDF) of blog documents under consideration are obtained and their Cosine Similarity score is computed. Synonyms are substituted against those unmatched keywords if the Cosine Similarity score so computed is below the threshold and an improved Cosine Similarity score of those documents under consideration is obtained. Computing the Influence Score after Synonym substitution (ISaS) of those bloggers under conflict, the top most influential blogger is profiled. The simulation results demonstrate that the proposed Profiling Top Most Influential Blogger using Synonym Substitution (PTMIBSS) algorithm is adequately accurate in determining the top most influential blogger at any instant of time considered.

Keywords

Blog Document, Content Analysis, Cosine Similarity Score, Influential Blogger, Profiling.

Full Text

     

D. Annapurna ¹, K. B. Raja ², K. R. Venugopal ², L. M. Patnaik ³

Affiliations
1 Department of Information Science and Engineering, PES Institute of Technology South Campus, Hosur Road, Bangalore 560100, IN
2 Department of Computer Science and Engineering, University Visvesvaraya College of Engineering, Bangalore University, Bangalore 560 001, IN
3 Indian Institute of Science, Bangalore, IN

Abstract

Hybrid protocol combines the advantages of proactive and reactive routing in adhoc network. The routing is initially established with some proactively prospected routes and then serves the demand from additionally activated nodes through reactive flooding. In this paper we propose A Quality Hybrid Service Discovery Protocol (QHSDP) for discovering services. A broadcast mechanism is used to get the service and routing information of the nodes present inside the zone. The routing and service information reduces the packet flooding in the network hence reducing collision and increasing packet delivery efficiency. Reduced control packets in turn reduces the battery power consumption. A query message is bordercasted through the peripheral nodes to the nodes outside the zone. This makes the discovery procedure more sclable, hence increasing the node's coverage and reducing the latency in the proposed technology compared to the existing technology.

Full Text

   

Deepthi K. Prasad ¹, L. Vibha ¹, K. R. Venugopal ²

Affiliations
1 Department of Computer Science Engineering, B N M Institute of Technology, IN
2 Department of Computer Science Engineering, University Visvesvaraya College of Engineering, IN

Abstract

Diabetic Retinopathy (DR) is complicated disorder in human retina which is affected due to an increasing amount of insulin in blood that results in vision impairment. Early detection of DR is used to support the patients to prevent blindness and to be aware of this disease. This paper proposes a novel technique for detecting DR using hybrid classifiers. It includes pre-processing of the image, segmentation of region of interest, feature extraction and classification. Retinal structures like microaneurysms, exudates, hemorrhages and blood vessels are segmented. Classification is performed with integration of Fuzzy logical System and Neural Network (NN) which improves the accuracy of classification. Experimentation is carried out with the MESSIDOR data set. Results are compared against various performance metrics like accuracy, sensitivity and specificity. An accuracy close to 100 percent and low average error rate of 0.012 are obtained using the proposed method. The results obtained are encouraging.

Keywords

Diabetic Retinopathy, Hybrid Classifier, Visual Impairment, Fundus Images, Classification, Fuzzy Neural Network.

Full Text

     

Deepthi K. Prasad ¹, L. Vibha ¹, K. R. Venugopal ²

Affiliations
1 Department of Computer Science and Engineering, BNM Institute of Technology, IN
2 University Visvesvaraya College of Engineering, Bangalore University, IN

Abstract

Glaucoma is one of the most complicated disorder in human eye that causes permanent vision loss gradually if not detect in early stage. It can damage the optic nerve without any symptoms and warnings. Different automated glaucoma detection systems were developed for analyzing glaucoma at early stage but lacked good accuracy of detection. This paper proposes a novel automated glaucoma detection system which effectively process with digital colour fundus images using hybrid classifiers. The proposed system concentrates on both Cup-to Disk Ratio (CDR) and different features to improve the accuracy of glaucoma. Morphological Hough Transform Algorithm (MHTA) is designed for optic disc segmentation. Intensity based elliptic curve method is used for separation of optic cup effectively. Further feature extraction and CDR value can be estimated. Finally, classification is performed with combination of Naive Bayes Classifier and K Nearest Neighbour (KNN). The proposed system is evaluated by using High Resolution Fundus (HRF) database which outperforms the earlier methods in literature in various performance metrics.

Keywords

Glaucoma, Optic Nerve, Cup-To-Disc Ratio, HRF Database, Hybrid Classifier.

Full Text