B-Scopus
Permanent URI for this communityhttps://dspace.psgrkcw.com/handle/123456789/3730
Browse
40 results
Search Results
Item EFFICIENT PREDICTION OF PHISHING WEBSITES USING SUPERVISED LEARNING ALGORITHMS (Conference Paper)(Elsevier B.V, 2012-12-09) Lakshmi V, Santhana; Vijaya, M SPhishing is one of the luring techniques used by phishing artist in the intention of exploiting the personal details of unsuspected users. Phishing website is a mock website that looks similar in appearance but different in destination. The unsuspected users post their data thinking that these websites come from trusted financial institutions. Several antiphishing techniques emerge continuously but phishers come with new technique by breaking all the antiphishing mechanisms. Hence there is a need for efficient mechanism for the prediction of phishing website. This paper employs Machine-learning technique for modelling the prediction task and supervised learning algorithms namely Multi layer perceptron, Decision tree induction and Naïve bayes classification are used for exploring the results. It has been observed that the decision tree classifier predicts the phishing website more accurately when comparing to other learning algorithms.Item SUPPORT VECTOR MACHINE BASED EPILEPSY PREDICTION USING TEXTURAL FEATURES OF MRI(Elsevier Ltd, 2010) Sujitha, V; Sivagami, P; Vijaya, M SEpilepsy is a disorder of the central nervous system, specifically the brain. It is a neurological malfunction affecting about 1% of the population and is the third most common neurological disorder following rheumatic heart disease and Alzheimer’s disease, but it imposes higher costs on society. Magnetic Resonance Imaging (MRI) is one of the most common diagnostic tests used for patients for epilepsy prediction. Shortage of radiologists and the large volume of MRI scan images that need to be analyzed may lead to labor intensive, expensive and inaccurate prediction. Hence there is a need to generate an efficient prediction model for making a correct diagnosis of epilepsy and accurate prediction of its type. This paper describes the modeling of epilepsy prediction using Support Vector Machines (SVM), a machine learning algorithm. The prediction model has been generated by training the support vector machine with descriptive features derived from MRI data of 350 patients and observed that the SVM based model with a Radial Basis Function (RBF) kernel produces 93.87% of prediction accuracy.Item CLASSIFICATION OF SEED COTTON YIELD BASED ON THE GROWTH STAGES OF COTTON CROP USING MACHINE LEARNING TECHNIQUES(IEEE Xplore, 2010-07-29) Jamuna, K S; Karpagavalli, S; Vijaya, M S; Revathi, P; Gokilavani, S; Madhiya, ECotton, popularly known as "White Gold" has been an important commercial crop of national significance due to the immense influence of its rural economy. Cotton seed is an important and critical link in the chain of agricultural activities extending farmer industry linkage. Cotton yield is associated with high quality seed as the seed contains in itself the blue print for the agrarian prosperity in incipient form. Transfer of technology to identify the quality of seeds is gaining importance. Hence this work employs machine learning approach to classify the quality of seeds based on the different growth stages of the cotton crop. Machine learning techniques - Naïve Bayes Classifier, Decision Tree Classifier and Multilayer Perceptron were applied for training the model. Features are extracted from a set of 900 records of different categories to facilitate training and implementation. The performance of the model was evaluated using 10 -fold cross validation. The results obtained show that Decision Tree Classifier and Multilayer Perceptron provides the same accuracy in classifying the seed cotton yield. The time taken to build the model is higher in Multilayer Perceptron as compared to the Decision Tree Classifier.Item ELECTROENCEPHALOGRAM WAVE SIGNAL ANALYSIS AND EPILEPTIC SEIZURE PREDICTION USING SUPERVISED CLASSIFICATION APPROACH(ACM Digital Library, 2010-09-16) Devi S. T, Pavithra; Vijaya, M SThe transient and unexpected electrical disturbances of the brain results in an acute disease called Epileptic seizures. A significant way for identifying and analyzing epileptic seizure activity in human is by using electroencephalogram (EEG) signal. Manually reviewing and analyzing lengthy data of EEG recordings, for detection and classification of electro graphical patterns at present requires trained personnel and time consuming. Hence, there is a need for an efficient automated system based on pattern classification for analysis and classification of seizure-related EEG signals to assist the expert in the diagnosis. This paper presents the modeling of epileptic seizure prediction as binary classification problem and provides a suitable solution by implementing supervised classification algorithms, namely Decision table, Naive Baye's Tree, k-NN and support vector machine. The classification models are trained using the EEG data sets and the prediction accuracy of the classifier has been evaluated using 10-fold cross validation. It has been observed that the model produce about 86% of prediction accuracy in predicting the presence of epileptic seizure in human brain.Item DEEP NEURAL NETWORKS FOR AFFINITY PREDICTION OF SPINOCEREBELLAR ATAXIA USING PROTEIN STRUCTURES(Journal of Advanced Research in Dynamical and Control Systems, 2019) Asha, P R; Vijaya, M SDrug identification for exceptional hereditary disorder like spinocerebellar ataxia (SCA) is defy and an obligatory chore in biomedical investigate. There are numeral paths obtainable for affinity prediction through diverse scores and features in a typical computational scaffold. However there is a need for creative approaches to further depict the affinity of spinocerebellar ataxia which will enable better prediction for drug identification. This research work depicts the importance of hotspots identification for protein-protein interaction and also rigid docking. The power of Deep Neural Networks is utilized to predict binding affinity patterns through 3d protein structures and interactive properties of the interacted complex that can be used for predicting binding affinity. The work is carried out with 626 protein structures to perform protein-protein interaction. The complexes attained from the protein-protein interaction are 313 and from these complexes features are extracted. Features like physio-chemical properties, energy calculations, interfacial and non-interfacial properties are extracted from the complexes to model the binding affinity and a dataset with 313 instances is developed. Deep learning architectures learn complicated patterns, by gradually building from simpler ones. Deep learning models has many layers, when it goes deeper the model gets refined to provide better results. Input is given as feature vectors for the reason that of the docking process. Deep neural network works well for prediction by learning the features and the signals between them and the experiments discovered the dominance of deep learning neural network when compared to traditional ensemble learning.Item IDENTIFICATION OF AUTISM SPECTRUM DISORDER USING A MULTI-LABEL APPROACH(Journal of Advanced Research in Dynamical and Control Systems, 2019) Sudha, V P; Vijaya, M SThere has been an increase in the application of Multi-dimensional models in domains like bioinformatics, image processing, video and audio processing and text categorization. Multi-dimensional approach has its advantages with regard to predictive accuracy and time taken to build the model. The search for candidate genes of Autism spectrum disorder (ASD) is complicated as it involves significant interactions among mutations in several genes. This work investigates multi- dimensional learning which builds a model to predict ASD related multiple variables simultaneously using varied features. The study explored the different methods of multi- dimensional learning. ASD related gene sequences were analysed using different characteristics and each sequence was represented by a profile of 58 features from three different categories specific to gene, substitution matrix and amino acid. By combining three different problem transformation approaches with three base classifiers and using two algorithm adaptation methods a total of 15 different configurations were constructed. The different configurations were evaluated with multiple measurements including Hamming Loss, Hamming score, Zero One loss, Exact match, Accuracy, training and testing time. The results showed that the problem transformation algorithm Nearest Set replacement together with Naive Bayes classifier outperformed the other configurations with 93.4 % accuracy and Hamming loss of 0.06, 0.12 Zero one loss.Item TEMPERATURE CONTROLLED PSO ON OPTIMIZING THE DBN PARAMETERS FOR PHONEME CLASSIFICATION(Springer Link, 2019-01-10) Laxmi Sree, B R; Vijaya, M SSpeech recognition has become an essential component to communicate with the latest gadgets and machines in ease through speech. Phoneme classification model for phonemes in Tamil continuous speech is built here by exploring the power of deep belief network (DBN), a powerful neural network architecture that is capable of learning complex problems. But building an efficient DBN highly relies on several parameters like number of layers, number of neurons, connection weights and bias. The effect of increasing the number of layers in DBN for phoneme recognition has been studied in our previous experiments. In addition, a methodology which employed particle swarm optimization (PSO) or its variants second generation PSO (SGPSO) and new method PSO (NMPSO) for optimizing the connection weights and bias of the DBN for phoneme classification were studied in our earlier work. Pre-training DBN with PSO faced the problem of particle stagnation and took longer time to converge, whereas DBN with SGPSO, NMPSO converges faster but still suffers from particle stagnation which prevents it from reaching an optimal solution. Here we try to minimize stagnation of particles in the population in addition to faster convergence by proposing a new improved PSO, named Temperature controlled TPSO to optimize the initial connection weights and bias parameters that controls the DBN efficiency. TPSO seems to converge faster with better optimizing the DBN connection weights and bias parameters when compared to the existing ones with reduced stagnation of population. The TPSO–DBN is designed and applied on a phoneme classification problem for Tamil continuous speech and found to classify phonemes comparatively better with a classification accuracy of 89.2%.Item AFFINITY PREDICTION OF SPINOCEREBELLAR ATAXIA USING PROTEIN-LIGAND AND PROTEIN-PROTEIN INTERACTIONS WITH FUNCTIONAL DEEP LEARNING(Blue Eyes Intelligence Engineering & Sciences Publication, 2019-06) Asha, P R; Vijaya, M SDrug discovery of incomparable hereditary disorder like spinocerebellar ataxia is confronted and an enforce task in biomedical study. There are number of paths available for affinity prediction through scoring functions and ideals in the catalog. Nevertheless there is a need for artistic access in portraying the affinity of spinocerebellar ataxia which will facilitate enhanced prediction for drug discovery. This research work portrays the significance of docking for protein-ligand interaction and protein-protein interaction with modeling through deep learning. Deep Neural Networks is utilized in predicting binding affinity with 3d protein structures and ligand. Predictive models have been built with features related to for protein-ligand interaction and protein-protein interaction. In the first case, 17 protein structures and 18 ligands were used. Each protein structure is docked with ligand to get essential features like energy calculations, properties of protein and ligand for predicting binding affinity. In the next case, repeat mutation is induced manually with 17 protein structures and docked with 18 ligands. To train the model, well-defined descriptors are squeezed from the docked complex. Third case employs protein-protein interaction of total of 626 protein structures and the complexes attained from the protein-protein interaction are 313. Features like energy calculations, physio-chemical properties and interfacial and non-interfacial properties are extracted for learning this model. Deep learning has the property of representation learning from the user defined features, which helps in accurate prediction of binding affinity. The predictive models are developed with functional deep neural network and their performances are compared with sequential deep neural network. Functional deep neural network have more flexibility to define layers, complements sequential deep neural network which results in improved performance.Item AFFINITY PREDICTION USING MUTATED PROTEIN-LIGAND DOCKING WITH REGRESSION TECHNIQUES OF SCA(Blue Eyes Intelligence Engineering & Sciences Publication, 2019-07) Asha, P R; Vijaya, M SDrug discovery for rare genetic disorder like spinocerebellar ataxia is very complicated in biomedical research. Numerous approaches are available for drug design in clinical labs, but it is time consuming. There is a need for affinity prediction of spinocerebellar ataxia, which will help in facilitating the drug design. In this work, the proteins are mutated with the information available from HGMD database. The repeat mutations are induced manually, and that mutated proteins are docked with ligand. The model is trained with extricated features such as energy profiles, rf-score, autodock vina scores, cyscore and sequence descriptors. Regression techniques like linear, polynomial, ridge, SVM and neural network regression are implemented. The predictive models are built with various regression techniques and the predictive model implemented with support vector regression is compared with support vector regression kernel. Among all regression techniques, SVR performs well than the other regression models.Item ANFIS FOR TAMIL PHONEME CLASSIFICATION(2019-08) Laxmi Sree, B R; Vijaya, M SPhoneme recognition is an intricate problem lying under non-linear systems. Most research in this area revolve around try to model the pattern of features observed in the speech spectra with the use of Hidden Markov Models (HMM), various types of neural networks like deep recurrent neural networks, time delay neural networks, etc. for efficient phoneme recognition. In this paper, we study the effectiveness of the hybrid architecture, the Adaptive Neuro-Fuzzy Inference System (ANFIS) for capturing the spectral features of the speech signal to handle the problem of Phoneme Recognition. In spite of a wide range of research in this field, here we examine the power of ANFIS for least explored Tamil phoneme recognition problem. The experimental results have shown the ability of the model to learn the patterns associated with various phonetic classes, indicated with recognition improvement in terms of accuracy to its counterparts.