International Journals

Permanent URI for this collectionhttps://dspace.psgrkcw.com/handle/123456789/157

Browse

Now showing 1 - 20 of 65

ANALYZING DATAMINING ALGORITHMS USING CAR DATASET
(CiiT International Journal of Data Mining and Knowledge Engineering, 2009-09) R, Deepalakshmi; N, Radha
The “Car Manufacturing” sector occupies a prime position in the development of automobile industry. In this paper, a proposed data mining application in car manufacturing domain is explained and experimented. The datasets are retrieved from UCI Machine learning repository. The purpose of this paper is to establish a classifier that is much more reliable in classifications for future objects. The classifier should provide sophisticated prediction to indicate the car data for a new input instance with some attributes, such as car type, body-style, horsepower and fuel. Such analysis helps in providing car market with base for more accurate result for the future market. The physical characteristics of a car viz. aspiration, number of doors, body-style, normalized losses, car-type, drive wheels, engine-location, wheel-base, curb-weight, horse-power, bore, stroke, city-mpg, highway-mpg, price, engine size, etc., are considered to determine the performance of a car. Hence development of such a classifier, though a voluminous task, is immensely essential in car manufacturing realm. Machine learning techniques can help in the integration of computer-based systems in predicting the quality of car and to improve the efficiency of the system. The classification models were trained by using 214 datasets. The predicted values for the classifiers were evaluated using 10-fold cross validation and the results were compared.
A COMPARATIVE STUDY OF MACHINE LEARNING ALGORITHMS APPLIED TO PREDICTIVE DIABETES DATA
(CiiT International Journal of Data Mining and Knowledge Engineering, 2009-11-25) K, Sathiyakumari; V, Pream Sudha
Healthcare industry encompasses abundant data, which is increasing everyday. Conversely, tools for analyzing these records are incredibly less. Machine learning provides a lot of techniques for solving diagnostic problems in a variety of medical domains. Intelligent systems are able to learn from machine learning methods, when they are provided with a set of clinical cases as training set. This paper aims at a comparative study of widely used supervised classification algorithms – Naïve Bayes, Multi Layer Perceptrons, Logistic Model Trees, and Nearest Neighbor with Generalized Exemplars applied to predictive diabetes dataset. The machine learning algorithms used in this study are chosen for their representability and diversity. They are evaluated on the basis of their accuracy, learning time and error rates.
PERFORMANCE ANALYSIS OF DATA MINING ALGORITHMS FOR IODINE DEFICIENCY DISORDERS
(International Journal of Computer Science and Information Technology (IJCSIT), 2009-12) Rubya T; Deepa Lakshmi R; Radha N
Biomedical datasets pose a unique challenge for machine learning and datamining techniques in order to extract accurate, comprehensible and hidden knowledge. This paper comprehensively investigates the role of a biomedical dataset such as hypothyroid dataset on the classification accuracy of an algorithm. The datasets are retrieved from UCI machine learning repository, Hypothyroid is a kind of disease, which occurs due to the insufficient production of thyroid hormones to the thyroid gland. This data classification is based on machine learning algorithms to provide very accurate predictions for real-world datasets and also to quantify the complexity of a biomedical dataset in terms of its missing values, imbalance ratio, and information gain from that dataset. The data mining classification algorithms in Weka tool is used to classify the data. The predicted values for the classifiers were evaluated using 10-fold cross validation and the results were compared.
A COMPARATIVE STUDY OF DIMENSION REDUCTION TECHNIQUES FOR CONTENT-BASED IMAGE RETRIEVAL
(The International Journal of Multimedia & Its Applications, 2010) Sasikala G; Kowsalya R; Punithavalli M
Efficient and effective retrieval techniques of images are desired because of the explosive growth of digital images. Content-based image retrieval is a promising approach because of its automatic indexing and retrieval based on their semantic features and visual appearance. This paper discusses the method for dimensionality reduction called Maximum Margin Projection (MMP). MMP aims at maximizing the margin between positive and negative sample at each neighborhood. It is designed for discovering the local manifold structure. Therefore, MMP is likely to be more suitable for image retrieval systems, where nearest neighbor search is usually involved. The performance of these approaches is measured by a user evaluation. It is found that the MMP based technique provides more functionalities and capabilities to support the features of information seeking behavior and produces better performance in searching images.
PREDICTION OF THE COMPRESSIVE STRENGTH OF HIGH PERFORMANCE CONCRETE MIX USING TREE BASED MODELING
(International Journal of Computer Applications, 2010) C, Deepa; V, Pream Sudha; K, SathiyaKumari
Concrete is the safest and sustainable construction material which is most widely used in the world as it provides superior fire resistance, gains strength over time and gives an extremely long service life. Its annual consumption is estimated between 21 and 31 billion tones. Designing a concrete mix involves the process of selecting suitable ingredients of concrete and determining their relative amounts with the objective of producing a concrete of the required, strength, durability, and workability as economically as possible. According to the National Council for Cement and Building Materials (NCBM), New Delhi, the compressive strength of concrete is governed generally, by the water-cement ratio. The mineral admixtures like fly ash, ground granulated blast furnace, silica fume and fine aggregates also influence it. The main purpose of this paper is to predict the compressive strength of the high performance concrete by using classification algorithms like Multilayer Perceptron, M5P Tree models and Linear Regression. The result from this study suggests that tree based models perform remarkably well in predicting the compressive strength of the concrete mix.
A TREE BASED MODEL FOR HIGH PERFORMANCE CONCRETE MIX DESIGN
(International Journal of Engineering Science and Technology, 2010) C, Deepa; K, Sathiya Kumari; V, Pream Sudha
Concrete is the sustainable construction material, which is most widely used in the world as it provides superior fire resistance, gains strength over time and gives an extremely long service life. Its annual consumption is estimated between 21 and 31 billion tones. The paper is aimed at guiding the selection of available materials and proportioning them as to produce the most economical concrete suitable for the desired purpose. According to the National Council for Cement and Building Materials (NCBM), New Delhi, the compressive strength of concrete is governed generally, by the water-cement ratio. The mineral admixtures like fly ash, ground granulated blast furnace, silica fume and fine aggregates also influence it. The main purpose of this paper is to find the accuracy for the compressive strength of high performance concrete by using classification algorithms like Multilayer Perceptron, Rnd tree models and C-RT regression. The result from this study suggests that tree based models perform remarkably well for designing the concrete mix.
A STUDY ON BIOMETRIC TEMPLATE SECURITY
(ICTACT Journal on Soft Computing, 2010-01) N, Radha; S, Karthikeyan
The increasing popularity of biometrics and cryptography is driven by the widespread stipulation on information security. Abundant efforts have been made in developing successful methods in these areas in order to accomplish an enhanced level of information security. There are two dominant issues in information security enhancement. One is to defend the user ownership and control the access to information by authenticating an individual’s identity. The other is to make sure the privacy and integrity of information and to secure communication. Cryptography is the science of writing in secret code. Secret-key cryptography and public-key cryptography are the two most important cryptographic architectures. The security of a cryptographic system is reliant on the secrecy of the cryptographic key. Biometric authentication or simply biometrics refers to establishing automatic personal recognition based on the physical and behavioral characteristics of an individual (e.g. face, voice, fingerprint, gait, hand geometry, iris, gene, etc.). Biometrics offers superior security and easier than traditional identity authentication systems (based on passwords and cryptographic keys).Since biometrics characteristics are naturally related with a particular individual, making them insusceptible to being stolen, forgotten, lost or attached. This paper presents a survey on various techniques proposed earlier in developing an authentication system for ensuring individual’s information security by combining biometric characteristics of that particular individual and the cryptographic techniques. In addition, it provides some fundamental idea for future research that may help in eliminating the problems associated with the present authentication systems
SECURING RETINA FUZZY VAULT SYSTEM USING SOFT BIOMETRICS
(Global Journal of Computer Science and Technology, 2010-09) N, Radha; S, Karthikeyan; P, Anupriya
The major concern of almost all the verification system is user authentication and security. This necessitates the development of a mechanism that ensures user security and privacy. A lot of research has been carried on this developing field and numerous techniques have been proposed earlier in literature. These traditional methods use tokens and passwords to provide security to the users. Uncertainly, it can be easily compromised by attackers and therefore it is significant to design verification system that ensures authentication. In recent years, technology has turned in favor of combining soft biometrics and cryptographic key generation technique. The principal feature of using soft biometric template is that it cannot be easily revoked by any unauthorized user. Most commonly used soft biometric features are iris, retina, face, fingerprint, voice and so on. Fuzzy vault is the framework which comprises of the combination of soft biometrics and cryptographic key generation technique. This fuzzy vault acts as an additional layer of security. This overcomes the limitation met by a biometrics system when implemented individually. This paper proposes a biometric verification system investigating the combined usage of soft biometrics features hardened by fuzzy vault scheme. This approach uses retina as a soft biometric since it is capable of providing best results. Experiments were conducted to investigate the performance of the proposed authentication system in ensuring the user security and privacy.
ACUTE CYSTITIS AND ACUTE NEPHRITIS PREDICTION USING MACHINE LEARNING TECHNIQUES
(Global Journal of Computer Science and Technology, 2010-09) Kowsalya R; Sasikala G; Sangeetha Priya J
Urinary System includes kidneys, bladder, ureters and urethra. This is the major system involves electrolyte balance of the body and filters the blood and excretes the waste products in the form urine. Even the small disturbance in the renal function will step in a disasters manifestation. Among them we are considering the two diseases that affect the system are acute cystitis and acute nephritis. This paper presents the implementation of three supervised learning algorithms, ZeroR, J48 and Naive Bayes in WEKA environment. The classification models were trained using the data collected from 120 patients. The trained models were then used for predicting the acute cystitis or acute nephritis of the patients. The prediction accuracy of the classifiers was evaluated using 10-fold cross validation and the results were compared.
PALMPRINT AND IRIS BASED AUTHENTICATION AND SECURE KEY EXCHANGE AGAINST DICTIONARY ATTACKS
(International Journal of Computer Applications, 2010-12) Tamilselvi P; Radha N
The Multimodal Biometric based user authentication systems are highly secured and efficient to use and place total trust on the authentication server where biometric verification data are stored in a central database. Such systems are prone to dictionary attacks initiated at the server side. In this paper, we propose an efficient approach based on multimodal biometrics (Palmprint and Iris) based user authentication and key exchange system. In this system, texture properties are extracted from the palmprint and iris images are stored as encrypted binary template in the server’s database, to overcome the dictionary attacks mounted by the server. The image processing techniques are used to extract a biometric measurement from the palmprint and iris. During login procedure the mutual authentication is done between the server and user and a symmetric key is generated on both sides, which could be used for further secure communication between them. Thus meet-in-the middle attack that happens between the user and the server can also be overcome. This system can be directly applied to strengthen existing password or biometric based systems without requiring additional computation.
MULTIMODAL BIOMETRICS BASED AUTHENTICATION AGAINST DICTIONARY ATTACKS
((IJCSE) International Journal on Computer Science and Engineering, 2010-12) Tamilselvi P; Radha N
The Multimodal Biometric based user authentication systems are highly secured and efficient to use and place total trust on the authentication server where biometric verification data are stored in a central database. Such systems are, prone to dictionary attacks initiated at the server side. In this paper, we propose an efficient approach based on multimodal biometrics (Iris and fingerprint) based user authentication and key exchange system. In this system, minutiae points and texture properties are extracted from the fingerprint and iris images are stored in the encrypted form in the server’s database, to overcome the dictionary attacks mounted by the server. The image processing techniques are used to extract a biometric measurement from the fingerprint and iris. During login procedure the mutual authentication is done between the server and user and a symmetric key is generated on both sides, which could be used for further secure communication between them. Thus meet-in-the middle attack that happens between the user and the server can also be overcome. This system can be directly applied to strengthen existing password or biometric based systems without requiring additional computation.
A SURVEY ON VARIOUS APPROACHES IN DOCUMENT CLUSTERING
(International Journal of Computer Technology and Applications, 2011) G, Manimekalai; K, Sathiyakumari; V, Preamsudha
Document clustering is the process of segmenting a particular collection of texts into subgroups including content based similar ones. The purpose of document clustering is to meet human interests in information searching and understanding. Nowadays all paper documents are in electronic form, because of quick access and smaller storage. So, it is a major issue to retrieve relevant documents from the larger database. Text mining is not a standalone task that human analysts typically engage in. The goal is to transform text composed of everyday language in a structured, database format. In this way, heterogeneous documents are summarized and presented in a uniform manner. Among others, the challenging problems of document clustering are big volume, high dimensionality and complex semantics
UNSUPERVISED APPROACH FOR DOCUMENT CLUSTERING USING MODIFIED FUZZY C MEAN ALGORITHM
(International Journal of Computer & Organization Trends, 2011) G, Manimekalai; V, Preamsudha; K, Sathiyakumari
Clustering is one the main area in data mining literature. There are various algorithms for clustering. There are several clustering approaches available in the literature to cluster the document. But most of the existing clustering techniques suffer from a wide range of limitations. The existing clustering approaches face the issues like practical applicability, very less accuracy, more classification time etc. In recent times, inclusion of fuzzy logic in clustering results in better clustering results. One of the widely used fuzzy logic based clustering is Fuzzy C-Means (FCM) Clustering. In order to further improve the performance of clustering, this thesis uses Modified Fuzzy C-Means (MFCM) Clustering. Before clustering, the documents are ranked using Term Frequency–Inverse Document Frequency (TF–IDF) technique. From the experimental results, it can be observed that the proposed technique results in better clustering results when compared to the existing technique
FUSION OF IRIS AND RETINA USING RANK-LEVEL FUSION APPROACH
(International Journal of Computer Technology and Application, 2011) Kavitha A; Radha N
Personal identification and authentication is difficulty in all the systems. Shared secrets like Personal Identification Numbers or Passwords and key devices such as Smart Cards are not presently sufficient in few situations. These traditional tokens-based systems may be easily stolen or lost. Biometrics is the only way of improving the capability to recognize the persons according to the physiological or behavioral features. In many real-world applications, unimodal biometric system suffers from some limitations of noise in sensed data, intra-class variation, inter-class similarities, non-universality and spoof attacks. Multibiometric systems seek to alleviate some of these problems by consolidating the evidence obtained from different sources. These systems help to achieve an increase in performance. This paper focused on developing a multimodal biometrics system, which uses biometrics such as iris and retina. Fusion of biometrics is performed by means of rank level fusion. The ranks of individual matchers are integrated using the borda count, and logistic regression approaches. The developed multimodal biometric system utilizes and Fisher’s Linear Discriminant (FLD) and Principal Component Analysis (PCA) methods for individual matchers (Iris and Retina) identification. The features from the biometrics are obtained by using the Fisher face. The experimental result shows the performance of the proposed multimodal biometrics system.
MACHINE LEARNING APPROACH FOR TAXATION ANALYSIS USING CLASSIFICATION TECHNIQUES.
(International Journal of Computer Applications, 2011-01) Deepalakshmi R; Radha N
Data mining process discovers useful information from the hidden data, which can be used for future prediction. Machine learning provides methods, techniques and tools, which help to learn automatically and to make accurate predictions based on past observations. The data are retrieved from the real time environmental setup. Machine learning techniques can help in the integration of computer-based systems in predicting the dataset and to improve the efficiency of the system. The main purpose of this paper is to provide a comparison of some commonly employed classification algorithms under same conditions. Such comparison helps to provide the accurate result in algorithms. Hence comparing the algorithms for such a classifier is a tedious task, for real time dataset. The classification models were experimented by using 365 datasets with 24 attributes. The predicted values for the classifiers were evaluated and the results were compared
SECURING RETINAL TEMPLATE USING QUASIGROUPS
(Special Issue of Security Issues and Solutions for Information, Computer and Networks, Journal of Advances in Information Technology (JAIT), 2011-05) Radha N; Rubya T; Karthikeyan S
Biometric plays important role in person recognition and identification. It is more secure than the traditional systems like password and token-based systems. The traditional systems can be stolen, misplaced or destroyed. But the biometric systems are based on the physiological or behavioral traits of the individual human being. It cannot be altered or misused by the unauthorized persons. It is more reliable than the traditional systems. Although it is powerful, immutable to vulnerable attacks So it is necessary to secure the biometric template using more efficient techniques. Cryptography is the most powerful technique to avoid vulnerable attacks. Recently it was found that the non-associative property of quasigroup helps achieving better security by having a randomly generated key for encr1yption. The operations involved in Quasigroup are computationally simple and can be efficiently used for protection of voluminous media like images, audio, video and different forms of multimedia in this paper the Quasigroup technique is used to provide better security to Retina Template
AN EFFICIENT SECURE BIOMETRIC SYSTEM USING NONINVERTIBLE GABOR TRANSFORM
(IJCSI International Journal of Computer Science, 2011-09) Radha N; Karthikeyan S
Biometric scheme are being widely employed because their security merits over the earlier authentication system based on records that can be easily lost, guessed or forged. High scale employments and the related template storage have increased the requirement to guard the biometric data stored in the system. Theft of biometric information is a negotiation of the user’s privacy. In Addition, the stolen biometric information can be used to access other biometric systems that have the similar feature provided for the user. Several alternative functions have been identified in literature for creating revocable or noninvertible biometric templates. Although, their security examination either disregards the distribution of biometric features or uses inefficient feature matching. This generally shows the way to unrealistic approximation of security. In this paper a novel approach for the Non-Invertible biometric system is proposed to secure the biometric template. Security of a feature transformation method can be assessed according to the two main factors: i) non-invertibility, and ii) diversity. Non-invertibility represents the complexity in obtaining the original biometric when the secure template is provided and diversity represents the complexity in guessing one secure template when a different secure template created from the identical biometric is provided. The proposed Non-invertible Gabor transform possess both the non-invertible and diversity features which enhances the security of the system to a large extent. The proposed approach is very much resistant to minor translation error and rotation distortion. The experimental result shows the better performance of the proposed technique compared to the existing system
A COMPARATIVE STUDY OF FUZZY MODELS IN DOCUMENT CLUSTERING
(International Journal on Computer Science and Engineering, 2012) G, Manimekalai; K, Sathiyakumari; V, Preamsudha
The availability of large quantity of text documents from the World Wide Web and business document management systems has made the dynamic separation of texts into new categories as a very important task for every business intelligence systems. Text document clustering is one of the emerging and most needed clustering techniques used to cluster documents with regard to similarity among documents. It is used widely in digital library management system in the modern context. Document clustering is widely applicable in areas such as search engines, web mining, information retrieval, and topological analysis. There are several clustering approaches available in the literature to cluster the document. But most of the existing clustering techniques suffer from a wide range of limitations. The existing clustering approaches face the issues like practical applicability, very less accuracy, more classification time etc. Thus a novel approach is needed for providing significant accuracy with less classification time. In recent times, inclusion of fuzzy logic in clustering provides better clustering results. One of the widely used fuzzy logic based clustering is Fuzzy C-Means (FCM) Clustering. In order to further improve the performance of clustering, this thesis uses Modified Fuzzy C-Means (MFCM) Clustering. The documents are ranked using Term Frequency–Inverse Document Frequency (TF–IDF) technique. From the experimental results, it can be observed that the proposed technique results in better clustering when compared to the FCM clustering technique
RANK LEVEL FUSION USING FINGERPRINT AND IRIS BIOMETRICS
(International Journal of Computer Science and Engineering, 2012-01) Radha N; Kavitha A
Authentication of users is an essential and difficult to achieve in all systems. Shared secrets like Personal Identification Numbers (PIN) or Passwords and key devices such as Smart cards are not presently sufficient in few situations. The biometric improves the capability to recognize the persons. A biometric identification system is an automatic recognition system that recognizes a person based on the physiological (e.g., fingerprints, face, retina, iris, ear) or behavioral (e.g., gait, signature, voice) characteristics. In many real-world applications, unimodal biometric systems often face has significant limitations due to sensitivity to noise, intra class variability, data quality, non-universality, and other factors. Multimodal biometric systems overcome some of these limitations. Multimodal biometric system provides more accuracy when compared to unimodal biometric system. The main goal of multimodal biometric system is to develop the security system for the areas that require high level of security. The proposed system focused on developing a multimodal biometrics system, which uses biometrics such as fingerprint and iris. Fusion of biometrics is performed by means of rank level fusion. The features from the biometrics are obtained by using the FLD (Fisher Linear Discriminant). The experimental result shows the performance of the proposed multimodal biometrics system. In this paper, the decision is made using rank level fusion and the ranks of individual persons are calculated using the Borda count, and Logistic regression approaches.
CLASSIFICATION OF USER OPINIONS FROM TWEETS USING MACHINE LEARNING TECHNIQUES
(International Journal of Advanced Research in Computer Science and Software Engineering (IJARCSSE), 2013-07) Poongodi S; Radha N
Online Social Network is a standard platform for collaboration, communication where people are connected to each other for sharing their opinion. In general, opinions can be articulated about anything like products, surveys, topics, individuals, organizations and events. There are two main types of textual information in web like facts and opinions. Facts can be expressed in defined terms by the user implicitly. To mine opinion, from the user defined facts is intellectually very demanding. User opinion is valuable data, which can be used for marketing research in business during decision making process. So opinion mining and classification plays a vital role in predicting what people think about products. In this work, basic Natural Language Processing (NLP) techniques and hash tag segments, emoticons are used for classification. The performance comparison of Support Vector Machine (SVM), Naïve bayes (NB) and Multilayer Perceptron (MLP) are done using weka. It is observed that the MLP gives better accuracy to classify the opinion from tweets