Conference Papers
Permanent URI for this collectionhttps://idr.nitk.ac.in/handle/123456789/28506
Browse
75 results
Search Results
Item An E-Learning System with Multifacial Emotion Recognition Using Supervised Machine Learning(Institute of Electrical and Electronics Engineers Inc., 2016) Ashwin, T.S.; Jose, J.; Raghu, G.; Guddeti, G.R.E-Learning systems based on Affective computingare popularly used for emotional/behavioral analysis of the users. Emotions expressed by the user is depicted by detecting the facialexpression of the user and accordingly the teaching strategies willbe changed. The present eLearning systems mainly focus on thesingle user face detection. Hence, in this paper, we proposemultiuser face detection based eLearning system using supportvector machine based supervised machine learning technique. Experimental results demonstrate that the proposed systemprovides the accuracy of 89% to 100% w.r.t different datasets(LFW, FDDB, and YFD). Further, to improve the speed ofemotional feature processing, we used GPU along with the CPUand thereby achieve a speedup factor of 2. © 2015 IEEE.Item A parallel dynamic programming approach for data analysis(Institute of Electrical and Electronics Engineers Inc., 2016) Deepak, A.; Shravya, K.S.; Chandrasekaran, K.In spite of presence of many classical and modified data analysis techniques, data analysis in the field of software engineering still remains a challenge because of the presence of large number of both continuous and discreet explanatory variables judging the outcome of one and more than one dependant variables. Requirement for an efficient multivariate data analysis technique which fulfils the constraints associated with software data led to the design of OSR (optimized set reduction) which uses a greedy algorithm for data analysis using both the principles of machine learning and conventional statistics. With the incoming of big data and other increasing dimensions of data set, we, through this paper, try to propose a new algorithm, based on the similar lines of optimised set reduction, using its strength to extract subsets. As the current trend of programming demands an algorithm to execute in parallel, we also propose a modification to our algorithm for it to run in a multicore platform with good efficiency. © 2015 IEEE.Item Windows malware detection based on cuckoo sandbox generated report using machine learning algorithm(Institute of Electrical and Electronics Engineers Inc., 2016) Shiva Darshan, S.L.S.; M.a, M.A.A.; Jaidhar, C.D.Malicious software or malware has grown rapidly and many anti-malware defensive solutions have failed to detect the unknown malware since most of them rely on signature-based technique. This technique can detect a malware based on a pre-defined signature, which achieves poor performance when attempting to classify unseen malware with the capability to evade detection using various code obfuscation techniques. This growing evasion capability of new and unknown malwares needs to be countered by analyzing the malware dynamically in a sandbox environment, since the sandbox provides an isolated environment for analyzing the behavior of the malware. In this paper, the malware is executed on to the cuckoo sandbox to obtain its run-time behavior. At the end of the execution, the cuckoo sandbox reports the system calls invoked by the malware during execution. However, this report is in JSON format and has to be converted to MIST format to extract the system calls. The collected system calls are structured in the form of N-Grams, which help to build the classifier by using the Information Gain (IG) as a feature selection technique. A comprehensive experiment was conducted to perceive the best fit classifier among the chosen classifiers, including the Bayesian-Logistic-Regression, SPegasos, IB1, Bagging, Part, and J48 defined within the WEKA tool. From the experimental results, the overall best performance for all the selected top N-Grams such as 200, 400, and 600 goes to SPegasos with the highest accuracy, highest True Positive Rate (TPR), and lowest False Positive Rate (FPR). © 2016 IEEE.Item A personalized recommender system using Machine Learning based Sentiment Analysis over social data(Institute of Electrical and Electronics Engineers Inc., 2016) Ashok, M.; Rajanna, S.; Joshi, P.V.; Kamath S․, S.S.Social Media platforms are already an indispensable part of our daily lives. With its constant growth, it has contributed to superfluous, heterogeneous data which can be overwhelming due to its volume and velocity, thus limiting the availability of relevant and required information when a particular query is to be served. Hence, a need for personalized, fine-grained user preference-oriented framework for resolving this problem and also, to enhance user experience is increasingly felt. In this paper, we propose a such a social framework, which extracts user's reviews, comments of restaurants and points of interest such as events and locations, to personalize and rank suggestions based on user preferences. Machine Learning and Sentiment Analysis based techniques are used for further optimizing search query results. This provides the user with quicker and more relevant data, thus avoiding irrelevant data and providing much needed personalization. © 2016 IEEE.Item Saliency prediction for visual regions of interest with applications in advertising(Springer Verlag service@springer.de, 2017) Jain, S.; Kamath S․, S.S.Human visual fixations play a vital role in a plethora of genres, ranging from advertising design to human-computer interaction. Considering saliency in images thus brings significant merits to Computer Vision tasks dealing with human perception. Several classification models have been developed to incorporate various feature levels and estimate free eye-gazes. However, for real-time applications (Here, real-time applications refer to those that are time, and often resource-constrained, requiring speedy results. It does not imply on-line data analysis), the deep convolution neural networks are either difficult to deploy, given current hardware limitations or the proposed classifiers cannot effectively combine image semantics with low-level attributes. In this paper, we propose a novel neural network approach to predict human fixations, specifically aimed at advertisements. Such analysis significantly impacts the brand value and assists in audience measurement. A dataset containing 400 print ads across 21 successful brands was used to successfully evaluate the effectiveness of advertisements and their associated fixations, based on the proposed saliency prediction model. © Springer International Publishing AG 2017.Item Information gain score computation for N-grams using multiprocessing model(Institute of Electrical and Electronics Engineers Inc., 2017) Shiva Darshan, S.L.S.; M.a, M.A.A.; Jaidhar, C.D.Currently, the Internet faces serious threat from malwares, and its propagation may cause great havoc on computers and network security solutions. Several existing anti-malware defensive solutions detect known malware accurately. However, they fail to recognize unseen malware, since most of them rely on signature-based techniques, which are easily evadable using obfuscation or polymorphism technique. Therefore, there is immediate requirement of new techniques that can detect and classify the new malwares. In this context, heuristic analysis is found to be promising, since it is capable of detecting unknown malwares and new variants of current malwares. The N-Gram extraction technique is one such heuristic method commonly used in malware detection. Previous works have witnessed that shorter length N-Grams are easier to extract. In order to identify and remove noisy N-Grams, a popular Feature Selection Technique (FST), namely, Information Gain (IG), which computes score for each N-Gram (feature) in the dataset has been used in this work. N-Grams with the highest IG score are considered as best features, while the remaining N-Grams are neglected. The IG-FST (Information Gain-Feature Selection Technique) is computational resource demanding and takes time to generate IG scores for larger N-Gram datasets, if the processing is to be accomplished in the sequential mode. To address this issue, the present work presents a multiprocessing model that computes IG scores rapidly for larger N-Gram datasets. The proposed model has been designed, implemented, and compared with the sequential mode of IG score computation. The experimental results demonstrate that the proposed multiprocessing model performance is 80% faster than the sequential model of IG score computation. © 2017 IEEE.Item Smart parking - An integrated solution for an urban setting(Institute of Electrical and Electronics Engineers Inc., 2017) Dsouza, K.B.; Yousuff, S.Parking could become a nightmare on a busy day, in a city like Delhi (India), which has about 7.35 million cars, as per MORTH Barclays Research (2012). An average of seventeen minutes and considerable amount of fuel is wasted in an effort to find a parking spot every time. Additional stress is induced due to parking hassles starting from finding an empty parking spot to relocating the car later. We propose a system leveraging the latest technologies that will help motorists overcome their parking problems and at the same time, make managing a parking space easier and cost effective by automating the entire process right from pre-booking a parking slot to making the payment. Since most of the parking spaces are equipped with CCTV surveillance cameras, we intend to use them to detect the presence of cars and measure the availability of parking spots within a parking space using techniques like image processing and machine learning. In order to test the performance of the proposed system, a prototype of the system is built that mimics the working of an actual parking space excluding minute details. A prototype of the application is built that would aid the user in booking the slot and guide him/her back to the allocated parking slot. The proposed system is compared with the existing systems and the results show superiority of the proposed system in terms of parameters like reliability, scalability and installation cost. © 2017 IEEE.Item Domain-specific sentiment analysis approaches for code-mixed social network data(Institute of Electrical and Electronics Engineers Inc., 2017) Pravalika, A.; Oza, V.; Meghana, N.P.; Kamath S․, S.Sentiment Analysis is one of the prominent research fields in Natural Language Processing because of its widespread real-world applications. Customer preferences, options and experiences can be analyzed through social media, reviews, blogs and other online social networking site data. However, due to increasing informal usage of local languages in social media platforms, multi-lingual or code-mixed data is fast becoming a common occurrence. Mixed code is generated when users use more than a single language in social network comments. Such data presents a significant challenge for applications using sentiment analysis and is yet to be fully explored by researchers. Existing sentiment analysis methods applied to monolingual social data are not suitable for code-mixed data due to the inconsistency in the grammatical structure in these sentences. In this paper, a novel method focused on performing effective sentiment analysis of bilingual sentences written in Hindi and English is proposed, that takes into account linguistic code switching and the grammatical transitions between the two considered languages. Experimental evaluation using real-world, code-mixed datasets obtained from Facebook showed that the proposed approach achieved very good accuracy and was also efficient performance-wise. © 2017 IEEE.Item Gender Detection using Handwritten Signatures(Institute of Electrical and Electronics Engineers Inc., 2018) Mohit Reddy, J.; Guru Pradeep Reddy, T.; Mishra, S.; Mulimani, M.; Koolagudi, S.G.In this paper, a method is proposed which uses both Image Processing and Machine Learning techniques which detects the gender of a person using handwritten signature. A photograph of a handwritten signature is given as input to the model which then extracts different features like pen pressure, slant angle, count external and internal contours etc. The features extracted from multiple images in the dataset are used to train the model, which then predicts the output of a new input given to it. Our objective is to collect unbiased datasets from a set of people and feed those signatures to the model, carrying out the statistical analysis and calculating the accuracy of the algorithm after every signature classification. We have used Adaboost classifier which gave a cross-validation accuracy of 73.2% compared to other classifiers like Gradient Boosting Classifier, Random Forest Trees and Multi-Layer Perceptron which gave 73.2%, 63.2% and 59.6% accuracies respectively. Copy Right © INDIACom-2018.Item Performance Evaluation of Filter-based Feature Selection Techniques in Classifying Portable Executable Files(Elsevier B.V., 2018) Shiva Darshan, S.L.; Jaidhar, J.The dimensionality of the feature space exhibits a significant effect on the processing time and predictive performance of the Malware Detection Systems (MDS). Therefore, the selection of relevant features is crucial for the classification process. Feature Selection Technique (FST) is a prominent solution that effectively reduces the dimensionality of the feature space by identifying and neglecting noisy or irrelevant features from the original feature space. The significant features recommended by FST uplift the malware detection rate. This paper provides the performance analysis of four chosen filter-based FSTs and their impact on the classifier decision. FSTs such as Distinguishing Feature Selector (DFS), Mutual Information (MI), Categorical Proportional Difference (CPD), and Darmstadt Indexing Approach (DIA) have been used in this work and their efficiency has been evaluated using different datasets, various feature-length, classifiers, and success measures. The experimental results explicitly indicate that DFS and MI offer a competitive performance in terms of better detection accuracy and that the efficiency of the classifiers does not decline on both the balanced and unbalanced datasets. © 2018 The Authors. Published by Elsevier B.V.