Evolutionary algorithm based Feature selection toolbox
Team Size: 2+
Role: Coding and Development
Environment: MATLAB, C, C++
Duration: December 2016 – February 2017
O.S: Window 7
Description:
During this period, we developed a novel toolbox for feature selection. This toolbox uses 15 evolutionary algorithm and seven classifiers for the development of wrapper-based feature selection techniques. The use of such a common tool becomes even more urgent when the source code of many proposed algorithms has not been made publicly available. In this period, we developed a MATLAB platform toolbox with several widely used classification performance indicators which can be used for feature selection. With a user-friendly graphical user interface, toolbox enables users to easily compare several evolutionary algorithms at one time. This toolbox is being used by the other fellow research scholars for the performance comparison.
Development of Nature Inspired Optimization Toolbox for twenty-six benchmark problems
Team Size: 2+
Role: Coding and Development
Environment: MATLAB, C, C++
Duration: August 2016 – November 2017
O.S: Window 7
Description:
Optimization is one of the most crucial issues in commercial systems design and product advancement. The main objective of optimization algorithms is to improve the performance of manufacturing processes which in turn reduce the cost of production and products become more competitive. During this period, I developed a toolbox called NIT (Nature-Inspired Toolbox) for the global minimization of test suites on 26 benchmark optimization problems. This toolbox is being used by the other fellow research scholars for the performance comparison. The underlying method makes extensive use of state-of-the-art algorithms from meta-heuristics optimization techniques.
Modeling of autonomous software agents’ behavior to defend HTTP request attacks Team Size
Team Size: 2+
Role: Coding and Development
Environment: MATLAB, C, C++
Duration: December 2015 – July 2016
O.S: Window 7
Description:
During this period, we developed a novel toolbox for HTTP denial of service attack with unsupervised techniques. This toolbox uses seven clustering algorithm and seven cluster validity measures for the development of DDOS attack detection from the raw web log file. It was a state government-funded project, where I worked as a project fellow. In this period, we developed a MATLAB platform for clustering toolbox with several widely used performance indicators. With a user-friendly graphical user interface, toolbox enables users to easily compare several clustering algorithms
Freelanced Project Developer, Raipur
June 2012- November 2015Designation- Software Developer
Responsibilities and Roles: I developed large number of codes for numerous clients which are mostly belonged to research-based models. In the development, I solved many problems from various domains including image processing, steganography, machine learning, data mining, cryptography, and network simulation. During this phase, I conducted two-day workshop on MATLAB tools and its application in MATS University Raipur.
National Institute of Technology, Raipur
July 2008- May 2012
Designation - Assistant Professor (on Contract)Responsibilities and Roles: A Finnish Pioneer in the soft computing research field. Responsibilities included: 1) Led team of graduate students on a year-long project; 2) Taught various subjects including Problem-solving logic building using C, Object-oriented programming in C++, Data structure and Algorithms.
Anjanies Institute of Engg. and Mgmt, Raipur
July 2007- June 2008
Designation - Lecturer
Responsibilities and Roles: Worked as a teaching faculty with responsibilities included
1) Tutorial classes for computer fundamentals.
2) Programming lab sessions
3) Student mentorship and Research lab in-charge.
An Empirical Analysis of Instance-based Transfer Learning Approach on Protease Substrate Cleavage Sites Prediction
The computational method for the classification of protease cleavage sites is significantly important in the inhibitors and drug design techniques. Matrix Metalloproteases (MMP) are one such protease that has a crucial role in the disease process. However, the challenge in the computational prediction of MMPs substrate-cleavage persists due to the availability of very few experimentally verified sites. The objective of this work is to explore the cross-domain learning in the classification of protease substrate-cleavage sites, such that the lack of availability of the one domain cleavage sites can be furnished by the other available domain knowledge. To achieve this objective we employed the TrAdaBoost algorithm and its two variants Dynamic TrAdaBoost and Multisource TrAdaBoost on the MMPs dataset available at PROSPER.
Proceedings: Accepted at MISP 2017 IIT Indore December 22- 24, 2017 (SCOPUS indexed AISC series of Springer).
Genetic Algorithm based Pre-processing strategy for High Dimensional Microarray Gene Classification.
Description: The objective of this work is to explore the possibilities of performance improvement in the large dimensional biomedical data with the joint alliance of machine learning and evolutionary algorithms for the design of effective healthcare systems. To accomplish the goal, the model targeted the preprocessing phase and developed framework based on Fisher Markov feature selection and evolutionary based binary discretization (EBD) for microarray gene expression classification. Several experiments were conducted on publicly available microarray gene expression dataset comprises of 10 data including colon tumor, lung and prostate cancer.
Published: Book chapter on ‘Application of Nature Inspired Intelligence on Bioinformatics:’ IGI Global
Multi-Objective Evolutionary approach for the Performance Improvement of Learners by Ensembling Feature selection and Discretization Technique on Medical data.
Description: The objective of the work is to incorporate both discretization and feature reduction as an ensemble model that has a unified framework for performing feature selection and discretization optimally. Wrapper-based feature selection algorithm selects the optimal features and categorized these selected features into two blocks namely discretized and non-discretized blocks. The feature belongs to the discretized block will participate in the binary discretization while the second block features will not be discretized and will be used as it is. The choice of optimal feature selection is achieved through wrapper based method, and feature splitting is done by entropy-based discretization technique. Minimizing the error rate during the feature selection and maximizing the information gain while discretization makes the problem multi-objective. Selection of optimal features called as feature subset, and the categorization of discretized and non discretized features from the feature subset is governed by the Multi-objective genetic algorithm (NSGA-II). For the establishment and acceptability of the proposed ensemble model, the experiment is conducted on the fifteen medical datasets and the metric such as accuracy, precision, recall and F-measure are computed for the performance evaluation of the classifiers while mean, standard deviation and evaluation time were used for the evolutionary algorithm performance.
Published: Journal of Current Medical Imaging Reviews, Impact Factor: 0.308
Evolutionary-based Optimal Ensemble Classifiers for HIV-1 Protease Cleavage Sites Prediction.
Description: Several Computational intelligence methods have been evaluated to predict cleavage sites. However, due to inconsistent findings regarding the superiority of one classifier over another and the usefulness of encoding techniques in general, more research is needed to provide advance confidence in computational results. The success of an HIV cleavage prediction system depends heavily on two things: the classifier being used and the features encoding technique applied. For the HIV cleavage, the role of appropriate feature encoding has not been paid adequate importance. In this investigation, we use two novel ideas for HIV Cleavage prediction. First, we propose a novel ensemble technique that optimizes the search space of 228 formed by seven encoding techniques and four SVM kernels (7x4) with the use of genetic algorithm. The second is the utilization of area under receiver operating characteristics (AUC) as a fitness measure for the evaluation of optimal ensemble.
Published: Expert System with Applications (Elsevier), Impact Factor: 3.928.
Evolutionary Based Ensemble Framework for Realizing Transfer Learning in HIV-1 Protease Cleavage Sites Prediction.
Description: The scarcity of sufficient training data and handling diverse and complex data especially in the field of bioinformatics are common and major challenges that need considerable attention for the domain adaptation. To cope with these challenges, we initiated the cross-domain adaptation ability in HIV-1 protease cleavage site prediction problem and proposed a framework that works on the principle of feature-mapping localization. In this cross-domain adaptability framework, we use feature-mapping localization concept where multiple feature extraction techniques will be optimally coupled to the several local kernels of SVM such that it can explore the diverse features. Each individual kernel is mapped to the feature subset generated from the feature extraction technique, and final prediction can be obtained by aggregating all the local trained kernels. The optimal coupled sequence of features and kernels are produced by genetic algorithm in a reasonable amount of time. To validate the effectiveness of the model, Tests were conducted on the 4 HIV-1 protease datasets offered at UCI machine learning database.
Published: Applied Intelligence (Springer), Impact Factor: 1.905.
RESEARCH INTEREST
Area of research interests are in the general domain of machine learning with an emphasis on ensemble learning and transfer learning which are used to increase the classification robustness against known and unknown classes. Moreover, I am highly interested in Evolutionary computing, Data quality assurance, Data visualization, Pattern recognition, Image processing, and Transfer learning.
