Browsing by Author "Kumar, N."

Now showing 1 - 20 of 23

A neural network-based predictive decision model for customer retention in the telecommunication sector
(Elsevier Inc., 2024) Thangeda, R.; Kumar, N.; Majhi, R.
Acquiring a new customer is far more expensive than retaining a customer. Hence, customer retention is a key aspect of business for a firm to maintain and improve on its market share and profit. The paper analyses customer retention strategies by employing an artificial neural network-based decision model to a real-life dataset collected from 311 mobile service users in India. Seven linear and non-linear adaptive models are developed using features related to customer dissatisfaction (DSF), customer disloyalty (DLF) and customer churn (CF). Findings of this study suggest that non-linear models are most efficient in predicting customer churn, and both DSF and DLF variables significantly affect the retention strategy. Three groups of customers are discussed in this study in the order of least likelihood of churning to most likelihood. Finally, a priority matrix based on key performance indicators is proposed to help service providers target potential customers to retain. © 2024 The Authors
An Effective Diabetic Retinopathy Detection Using Hybrid Convolutional Neural Network Models
(Springer Science and Business Media Deutschland GmbH, 2023) Kumar, N.; Ahmed, R.; Venkatesh, B.H.; Anand Kumar, M.
Loss of vision in the present era of the developing world is mainly caused by diabetic retinopathy. More than 103 million people are believed to be affected. It is estimated that around 40 million beings have diabetes in the United States, and according to the World Health Organization (WHO), 347 million people are living with the disease globally. Diabetic retinopathy (DR) is a long-term diabetes-related eye condition. Roughly, 45–50% of the American citizens suffering from diabetes undergo some unique stages that can be categorized. When DR is diagnosed on a timely basis, the possibility of it extending to the course of vision impairment can be delayed and stopped, though this is not entirely true and a very daunting task because it seldom reveals any symptom before it escalates to a stage of no return to effectively treat it. The paper uses convolutional neural network models to achieve an effective classification for diabetic detection of retinal fundus images. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.
An Image Transmission Technique using Low-Cost Li-Fi Testbed
(Institute of Electrical and Electronics Engineers Inc., 2021) Salvi, S.; Geetha, V.; Maru, H.; Kumar, N.; Ahmed, R.
Visible Light Communication (VLC) or Light Fidelity (Li-Fi) with Light Emitting Diodes (LEDs) as transmitter and light sensor as receiver will turn the present lightening system into a communication system. Li-Fi based data communication provides secure communication within the luminous coverage of the light source. Thus, it has several applications in places where Radio Frequency interference is not desirable. Similar to other wireless communication techniques even Li-Fi is used for transmission and reception of digital data. Li-Fi system can also be used to transfer images from one device to another. In this paper, a preliminary study is discussed by proposing and implementing an encoding and decoding scheme for transmission of the binary image using Li-Fi. The proposed system is evaluated based on the light intensity, distance, accuracy, size of the image, image resolution, and transmission time. Â© 2021 IEEE.
Analysis and prediction of COVID-19 trajectory: A machine learning approach
(John Wiley and Sons Ltd, 2021) Majhi, R.; Thangeda, R.; Sugasi, R.P.; Kumar, N.
The outbreak of Coronavirus 2019 (COVID-19) has impacted everyday lives globally. The number of positive cases is growing and India is now one of the most affected countries. This paper builds predictive models that can predict the number of positive cases with higher accuracy. Regression-based, Decision tree-based, and Random forest-based models have been built on the data from China and are validated on India's sample. The model is found to be effective and will be able to predict the positive number of cases in the future with minimal error. The developed machine learning model can work in real-time and can effectively predict the number of positive cases. Key measures and suggestions have been put forward considering the effect of lockdown. © 2020 John Wiley & Sons Ltd
Continuous Sign Language Recognition Using Leap Motion Sensor
(Institute of Electrical and Electronics Engineers Inc., 2024) Kumar, N.; Ahmed, R.; Venkatesh, B.H.; Salvi, S.; Panjwani, Y.
A vital communication tool that connects persons with hearing and speech impairments worldwide is sign language. Sign language involves mostly hand movements plus face gestures, which are interpreted by recognizing these gestures to form meaningful sentences. In this study, we use two machine learning models: Long Short-Term Memory (LSTM) and Support Vector Machines (SVM), to predict signs. A dataset of 42 unique sign words and 28 sentences was used to train and evaluate our models. Our method uses depth sensors, like the Leap Motion gadget, to improve sign language recognition (SLR).Worldwide, sign language is an essential communication tool for people with speech and hearing impairments. Sign languages are primarily made up of hand gestures and facial expressions, and their meaning is communicated through precise gesture interpretation. Our models were trained on a dataset containing 42 distinct sign words and 28 sentences, achieving an accuracy of 90.35% for word prediction and 98.21 for sentence prediction. The LSTM model outperformed the SVM model, which had accuracies of 85.96% and 89.58% for words and sentences, respectively. By using depth sensors like the Leap Motion device, our approach aims to enhance sign language recognition (SLR). Â© 2024 IEEE.
Deep Neural Network Models for Detection of Arrhythmia based on Electrocardiogram Reports
(Institute of Electrical and Electronics Engineers Inc., 2020) Ghuge, S.; Kumar, N.; Shenoy, T.; Kamath Sâ€¤, S.
Electrocardiogram (ECG) is an indicative technique using which the heartbeat time series of a patient is recorded on the moving strip of paper or line on the screen, for irregularity analysis by experts, which is a time-consuming manual process. In this paper, we proposed a deep neural network for the automatic, real-time analysis of patient ECGs for arrhythmia detection. The experiments were performed on the ECG data available in the standard dataset, MIT-BID Arrhythmia database. The ECG signals were processed by applying denoising, detecting the peaks, and applying segmentation techniques, after which extraction of temporal features was performed and fed into a deep neural network for training. Experimental evaluation on a standard dataset, using the evaluation metrics accuracy, sensitivity, and specificity revealed that the proposed approach outperformed two state-of-the-art models with an improvement of 2-7% in accuracy and 11-16% in sensitivity. Â© 2020 IEEE.
Deepfake Image Detection using CNNs and Transfer Learning
(Institute of Electrical and Electronics Engineers Inc., 2021) Kumar, N.; Pranav, P.; Nirney, V.; Geetha, V.
Headways in deep learning has enabled the creation of fraudulent digital content with ease. This fraudulent digital content created is entirely indistinguishable from the original digital content. This close identicalness has what it takes to cause havoc. This fraudulent digital content, popularly known as deepfakes having the potential to change the truth and decay faith, can leave impressions on a large scale and even our daily lives. Deepfake is composed of two words, the first being deep: deep learning and the second being fake: fake digital content. Artificial intelligence forming the nucleus of any deepfake formulation technology empowers it to dodge most of the deepfake detection techniques through learning. This ability of deepfakes to learn and elude detection technologies is a matter of significant concern. In this research work, we focus on our efforts towards the detection of deepfake images. We follow two approaches for deepfake image detection, and the first is to build a custom CNN based deep learning network to detect deepfake images, and the second is to use the concept of transfer learning. Â© 2021 IEEE.
Electro-discharge machining of microholes on 3d printed Hastelloy using the novel tool-feeding approach
(KeAi Publishing Communications Ltd., 2025) Korgal, A.; Shettigar, A.K.; P, N.K.; Kumar, N.; Bindu Madhavi, B.M.
Hastelloy, a nickel-based superalloy renowned for its exceptional resistance to corrosion at high temperatures, is widely used in sectors such as nuclear, aerospace, chemical processing, and pharmaceuticals. Microelectrical discharge machining (?-EDM) is crucial for generating microholes and channels on Hastelloy. Since it effectively addresses difficulties like work hardening, high strength & wear resistance, and low thermal conductivity in traditional machining. Microholes play a major role in many critical components for precise control of fluids in fuel injectors, managing heat in turbine blades, controlled gas exchange, etc. The current research investigates the drilling of 8:1 aspect ratio microholes machined by 400 ?m diameter electrodes. This study investigated the influence of tool material (tungsten carbide, carbide drill bit, and brass) on ?-EDM performance. Compared to tungsten carbide and carbide drill bits, brass exhibited significantly lower electrode wear, leading to more precise microholes with reduced overcut and taper angle. However, brass also required a substantially longer machining time. Carbide drill bits offered a balance between wear resistance, machining time, and overcut/taper angle. © 2024 The Authors
Improved Variable Round Robin Scheduling Algorithm
(Institute of Electrical and Electronics Engineers Inc., 2021) Mangukia, A.; Ibrahim, M.; Golamudi, S.; Kumar, N.; Anand Kumar, M.
The scheduling strategy used has an impact on system efficiency. It assigns processes in a specified order in order to improve those functions. The Round Robin method is one of several types of scheduling algorithms. In Round Robin, each process is allotted a time quantum (TQ), which indicates that each process consumes the same amount of time as the other processes. There is no precedence among the functions, and the CPU has a relatively short response time. The Time Quantum value cannot be too small because the number of context shifts will increase and hamper performance, while a Time Quantum value that is too large will harm the ART. The suggested study focuses on making the Round Robin approach more practical and efficient. Â© 2021 IEEE.
Intelligent GD&T symbol detection in mechanical drawings: a comparative study of YOLOv11, Faster R-CNN, and RetinaNet for quality assurance
(Springer, 2025) Narendra Reddy, T.N.; Kumar, N.; Ponnappa, N.P.; Mohana, N.; Vinod, P.; Herbert, M.A.; Rao, S.S.
Geometric dimensioning and tolerancing (GD&T) symbols play a vital role in engineering drawings by specifying allowable variations in part geometry to ensure manufacturing precision and functional performance. Manual identification and extraction of these symbols is labour-intensive, prone to human error, and increasingly unsuitable for fast-paced production environments, as it significantly increases quality inspection time and indirectly delays overall product delivery. This research is specifically conducted to support the development of intelligent quality management systems by integrating machine learning algorithms capable of detecting GD&T symbols directly from CAD-generated mechanical drawings. Such capability is essential for automating inspection processes and enabling reliable data extraction from design files, which are foundational to digital manufacturing workflows. Additionally, with many commercial quality automation tools being prohibitively expensive for small and medium-sized enterprises (SMEs) and micro, small, and medium enterprises (MSMEs), there is a pressing need for cost-effective, indigenous solutions. This study addresses that gap by evaluating three state-of-the-art deep learning-based object detection models—YOLOv11, Faster R-CNN, and RetinaNet—for GD&T symbol recognition. Each model was trained on a custom dataset annotated with diverse GD&T symbols, and performance was assessed using standard evaluation metrics: accuracy, recall, F1 score, and inference speed. The results show that while all three models demonstrate robust performance, YOLOv11 strikes the best balance between detection accuracy and real-time execution. This comparative study not only guides R&D teams in selecting the most suitable model for quality automation tasks but also contributes to the broader goal of enabling affordable, scalable, and intelligent visual inspection systems for SMEs and MSMEs. © The Author(s) 2025.
Investigating the performance of electromagnetic pump fabricated using tool based micromachining setup for microdelivery of fluid
(Bangladesh University of Engineering and Technology, 2019) Veeresha, R.K.; Karegoudra, M.K.; Rao, M.; Rao, R.; Kumar, N.
Micropumps play an important role in the delivery of insulin, hormonal and pain management for biomedical application. The main reason for the micropump is to pump a small amount of fluid to the target area and also control the pumping fluid. Electromagnetically operated pump fabricated using tool-based micromachining setup for the micro-delivery of the fluid. The electromagnetic pump was supplied with an input voltage of 6V-12V and at different frequencies starting from 1Hz to 5Hz with the increment of 1Hz. The maximum head developed is at a frequency of 3Hz which the optimum frequency for this configuration of an electromagnetic pump is. The maximum head obtained at this optimal frequency is 25mm. Finally, in order to measure the flow rate of the electromagnetic pump the pump was actuated at 3Hz frequency alone by varying the head of the micropump from 4 to 20mm. © 2019 Bangladesh University of Engineering and Technology. All rights reserved.
Mechanical Characterization and Finite Element Analysis of Jute-Epoxy Composite
(2018) Sangamesh, R.; Kumar, N.; Ravishankar, K.S.; Kulkarni, S.M.
Natural fiber composite materials are such an appropriate material, that replaces synthetic composite materials for many of practical applications where we need high strength and low density. Natural fiber composites combine the technological, ecological and economical aspects. This leads to discovering its vast applications in the aeronautics, automotive, marine and sporting sectors. This paper deals with the study on mechanical characterization (Tensile, Compression and Flexural) of jute/epoxy (JE) polymer composite. The flexural properties of composites are experimentally tested and are simulated in commercially available FEA software. Flexural tested results are in good agreement with FEA results. Scanning electron microscopy (SEM) analysis of the failed samples reveals the matrix dominated failure. � The Authors, published by EDP Sciences, 2018.
Mechanical Characterization and Finite Element Analysis of Jute-Epoxy Composite
(EDP Sciences edps@edpsciences.com, 2018) Sangamesh, R.; Kumar, N.; Ravishankar, K.S.; Kulkarni, S.M.
Natural fiber composite materials are such an appropriate material, that replaces synthetic composite materials for many of practical applications where we need high strength and low density. Natural fiber composites combine the technological, ecological and economical aspects. This leads to discovering its vast applications in the aeronautics, automotive, marine and sporting sectors. This paper deals with the study on mechanical characterization (Tensile, Compression and Flexural) of jute/epoxy (JE) polymer composite. The flexural properties of composites are experimentally tested and are simulated in commercially available FEA software. Flexural tested results are in good agreement with FEA results. Scanning electron microscopy (SEM) analysis of the failed samples reveals the matrix dominated failure. Â© The Authors, published by EDP Sciences, 2018.
Molecular-InChI: Automated Recognition of Optical Chemical Structure
(Institute of Electrical and Electronics Engineers Inc., 2022) Kumar, N.; Rashmi, M.; Ramu, S.; Reddy Guddeti, R.M.
With the advent of a new era dominated by digital media and publications in recent years, the importance of striking a balance between traditional and new modes of operation has become increasingly apparent. It has been standard practice in the field of chemistry for decades to express chemical compounds using their structural forms, referred to as the Skeletal formula. In this research, we tried to interpret these old chemical structure images, extracted from old literature, to transform pictures back to the underlying chemical structure labeled as InChI text using a huge set of synthetic image data produced by Bristol-Myers Squibb. In this paper, we propose an improved synthetic data and an Encoder-Decoder-based deep learning-based model to automatically represent these molecular images into their underlying InChI representation. Â© 2022 IEEE.
Multilayer Technique to Secure Data Transfer in Private Cloud for SaaS Applications
(Institute of Electrical and Electronics Engineers Inc., 2020) Ghuge, S.; Kumar, N.; Savitha, S.; Suraj, V.
In recent times Cloud Computing[CC] has emanated as a substitute paradigm for hosting and providing services over the Internet. Software as a Service (Saas) is one among such services that deliver services to the end-users on pay-as-you-go manner. In spite of all its advantages, security always seems to be major drawback. For securing the users' data on the cloud, this paper proposes an application model for any SaaS application hosted on a private cloud environment. The application is divided into two micro-services, where the first one is Application Layer Firewall and second one is a secured application to login and send sensitive data. The application layer firewall checks for any malicious activity and prevents the intruder to access the features present in the application. Subsequently, a Hidden Markov Model layer is implemented which is a probability-based intrusion detection technique. The second micro-service uses Advanced Encryption Standard (AES) encryption algorithm to encrypt documents having sensitive data, which have to be transferred within the private cloud. Further security is provided by proposing a novel Video Steganography approach using the Least Significant Bit (LSB) technique. This paper gives a detailed structure of hiding the data using multiple levels of security. Thus, this paper provides a holistic approach to implement a high level of security in SaaS applications. Â© 2020 IEEE.
Numerical study on the effect of steel fibers on fracture and size effect in concrete beams
(Elsevier Ltd, 2023) Yadav, D.; Prashanth, M.H.; Kumar, N.
The construction sector uses concrete extensively all around the world. Concrete contains a lot of microcracks even before it is loaded. When a tensile force is applied, these microcracks attempt to open up. While designing, the strength of concrete in its tensile zone is ignored. The strength and ductility of the concrete can be improved due to the addition of steel fibers. Steel fibers use a bridge mechanism to restrict the micro-cracks spread. This study uses ABAQUS to numerically analyze the behaviour of the Steel Fiber Reinforced Concrete (SFRC) beams. Two grades of concrete are studied, M20 and M60, for varying volumetric percentages of steel fibers. It was observed from the study that the ultimate load increases by around 52% and 41% for M25 and M60 grade concrete, respectively, by adding 1% of steel fiber. Fracture properties such as fracture toughness and fracture energy are calculated. The addition of steel fibers enhanced fracture toughness and energy significantly. Adding 1% fiber increases fracture toughness by around 56% and 34% and fracture energy by around 169% and 136% for M25 and M60 concrete, respectively. The size effect on SFRC beams is studied to determine the size-independent fracture parameters. © 2023
Peer Consonance in Blockchain based Healthcare Application using AI-based Consensus Mechanism
(Institute of Electrical and Electronics Engineers Inc., 2020) Kumar, N.; Parangjothi, C.; Guru, S.; Manjappa, M.
The term 'Blockchain', commonly referred to as the brain behind the Bitcoin network, works on the simple principle of the presence of a distributed and decentralized ledger in a public or private network. Since blockchain is decentralized, it is the duty of the Consensus Algorithm to substantiate the details in the blockchain. Traditional consensus algorithms such as Proof of Work (PoW) and Proof of Stake (PoS), although widely used, are a matter of concern due to computationally expensive operations and convergence towards a monopolized system respectively. Though optimizations of PoW and PoS algorithms were subsequently introduced, their features precincts. This paper aims to provide a solution by presenting a consensus algorithm based on Artificial Intelligence (AI) technology while maintaining the fairness of the system. A Healthcare based system was set up on top of the blockchain network to generate the dataset about the miners in order to train our neural network. On the whole, it incorporated the advantages of the state of the art consensus models which can increase the efficiency of the healthcare industry while diminishing their demerits. Â© 2020 IEEE.
Poly(N,N-diethyl acrylamide)/functionalized graphene quantum dots hydrogels loaded with doxorubicin as a nano-drug carrier for metastatic lung cancer in mice
(Elsevier Ltd, 2019) Havanur, S.; Batish, I.; Cheruku, S.P.; Gourishetti, K.; JagadeeshBabu, J.; Kumar, N.
Cancer has emanated as a daunting menace to human-kind even though medicine, science, and technology has reached its zenith. Subsequent scarcity in the revelation of new drugs, the exigency of salvaging formerly discovered toxic drugs such as doxorubicin has emerged. The invention of drug carrier has made drug delivery imminent which is ascribable to its characteristic traits of specific targeting, effective response to stimuli and biocompatibility. In this paper, the nanoscale polymeric drug carrier poly(N,N-diethyl acrylamide) nanohydrogel has been synthesized by inverse emulsion polymerization. Lower critical solution temperature of the polymeric carrier has been modified using graphene quantum. The particle size of pure nanohydrogel was in the range of 47 to 59.5 nm, and graphene quantum dots incorporated nanohydrogels was in the range of 68.1 to 87.5 nm. Doxorubicin (hydroxyl derivative of anthracycline) release behavior as a function of time and temperature was analyzed, and the Lower critical solution temperature of the synthesized nanohydrogels has been found to be in the range of 28–42 °C. Doxorubicin release characteristics have improved significantly as the surrounding temperature of the release media was increased near to physiological temperature. Further, the cumulative release profile was fitted in the different kinetic model and found to follow a Fickian diffusion release mechanism. The hydrogel was assessed for its cytotoxicity in B16F10 cells by MTT assay. In-vivo studies were done to study the lung metastasis by melanoma cancer and the results showed a rational favorable prognosis which was confirmed by evaluating hematological parameters and the non-immunogenic nature of nanohydrogel by cytokine assay. Comprehensively, the results suggested that poly(N,N-diethyl acrylamide) nanohydrogels have potential application as an intelligent drug carrier for melanoma cancer. © 2019 Elsevier B.V.
Program Slicing Analysis with KLEE, DIVINE and Frama-C
(Institute of Electrical and Electronics Engineers Inc., 2021) Kumar, N.; Neema, S.; Das, M.; Mohan, B.R.
Optimizing the Time complexity of any program is still the most researched and sought area for researchers. At the industry level, the Software execution timing is the dominant criteria for Workload selection. One prominent method for reducing the Time complexity of a program is by using program slicing configuration, without affecting the program flow. Program slicing is the process of slicing a program in such a way that it reduces the time of debugging. This paper presents a timed-based analysis of a program with and without slicing with the help of different verification tools, namely KLEE, DIVINE, and Frama-C. This paper aims to compare these tools based on the timing of debugging and validity of a program before and after slicing. Â© 2021 Chinese Automation and Computing Society in the UK-CACSUK.
Sketch-Based Image Retrieval Using Convolutional Neural Networks Based on Feature Adaptation and Relevance Feedback
(Springer Science and Business Media Deutschland GmbH, 2022) Kumar, N.; Ahmed, R.; B Honnakasturi, V.; Kamath Sâ€¤, S.; Mayya, V.
Sketch-based Image Retrieval (SBIR) is an approach where natural images are retrieved according to the given input sketch query. SBIR has many applications, for example, searching for a product given the sketch pattern in digital catalogs, searching for missing people given their prominent features from a digital people photo repository etc. The main challenge involved in implementing such a system is the absence of semantic information in the sketch query. In this work, we propose a combination of image prepossessing and deep learning-based methods to tackle this issue. A binary image highlighting the edges in the natural image is obtained using Canny-Edge detection algorithm. The deep features were extracted by an ImageNet based CNN model. Cosine similarity and Euclidean distance measures are adopted to generate the rank list of candidate natural images. Relevance feedback using Rocchioâ€™s method is used to adapt the query of sketch images and feature weights according to relevant images and non-relevant images. During the experimental evaluation, the proposed approach achieved a Mean average precision (MAP) of 71.84%, promising performance in retrieving relevant images for the input query sketch images. Â© 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.