Qualification: 
M.Tech, B-Tech
ashaashok@am.amrita.edu

Asha Ashok currently works as an Assistant Professor (Sr. Gr.) at the Department of Computer Science and Engineering at Amrita School of Engineering, Amritapuri.

Publications

Publication Type: Conference Paper

Year of Publication Title

2017

A. Manghat and Asha Ashok, “Abnormality prediction in high dimensional dataset among semi supervised learning approaches”, in 2017 International Conference on Advances in Computing, Communications and Informatics (ICACCI), Udupi, India, 2017.[Abstract]


Abnormality or inconsistency detection within a data is an attempt to make a distinction between usual and exceptional data instances. In this paper, we have proposed a novel methodAbnormality Prediction in High Dimensional Dataset among Semi Supervised Learning approaches (AP-HDD-SSL) to match the efficiencies of different semi supervised machine learning approaches using high dimensional KDD CUP 99 dataset. The pre-processing phase with dimensionality diminution is done prior to clustering using RFE (Random Forest Ensemble). Clustering with k-Means is initiated after the pre-processing phase for storing the most anomalous cluster. The classification within the cluster is done with semi-supervised learning approaches: k-Nearest Neighbour (k-NN), Linear Discriminant Analysis (LDA), Support Vector Machine-RFE(SVM-RFE), that are analysed and compared with the existing Over Sampling-PCA(os-PCA) method. The comparison results with Pima Indian and KDD cup 99 in terms of Accuracy, Detection Rate and AUC scores summarizes that AP-HDD-SSL with SVM-RFE outranked the other approaches.

More »»

2016

Asha Ashok, Smitha, S., and Krishna, M. H. K., “Attribute reduction based anomaly detection scheme by clustering dependent oversampling PCA”, in Symposium on Emerging Topics in Computing and Communications (SETCAC’16), International Conference on Advances in Computing, Communications and Informatics (ICACCI), 2016.[Abstract]


Anomaly detection refers to the task of estimating and finding patterns which do not comply with the general behavior of data. Moreover, a range of assumptions are made so as to differentiate between normal and deviated data instances. This paper describes a solution approach to this problem using a two-step phase including an important preprocessing phase and anomaly detection phase. For the preprocessing phase, we have used two methods mainly: Recursive Feature Elimination method (RFE) and Random Forest Ensemble (RF-Ensemble) method. For the next phase of anomaly detection, we have used Clustering based Oversampling PCA (os-PCA) methodology. The k-median clustering approach is utilized for this purpose. The technique was implemented and tested on various standard data sets like Pima, Splice etc. The results were also compared with the existing state of the methods in this field like online Oversampling PCA, Naive Oversampling PCA, decremental PCA, Local Outlier Factor, Angle Based Outlier detection and Median Based Outlier Detection approaches. The testing results confirm that the proposed approach outperformed all other methods on the basis of accuracy, AUC scores etc.

More »»

Publication Type: Conference Proceedings

Year of Publication Title

2016

Prathibhamol CP and Asha Ashok, “Solving multi label problems with clustering and nearest neighbor by consideration of labels”, 2nd International Symposium on Signal Processing and Intelligent Recognition Systems (SIRS'15), Advances in Intelligent Systems and Computing, vol. 425. Springer , pp. 511-520, 2016.[Abstract]


In any Multi label classification problem, each instance is associated with multiple class labels. In this paper, we aim to predict the class labels of the test data accurately, using an improved multi label classification approach. This method is based on a framework that comprises an initial clustering phase followed by rule extraction using FP-Growth algorithm in label space. To predict the label of a new test data instance, this technique searches for the nearest cluster, thereby locating k-Nearest Neighbors within the corresponding cluster. The labels for the test instance are estimated by prior probabilities of the already predicted labels. Hence, by doing so, this scheme utilizes the advantages of the hybrid approach of both clustering and association rule mining.The proposed algorithm was tested on standard multi label datasets like yeast and scene. It achieved an overall accuracy of 81% when compared with scene dataset and a 68% in yeast dataset. © Springer International Publishing Switzerland 2016.

More »»
Faculty Research Interest: