Senthil Velan S
@cmrit.ac.in
Professor, Information Science and Engineering
CMR Institute of Technology, Bengaluru, India
RESEARCH INTERESTS
Software Engineering, Data Analytics, Machine Learning, Artificial Intelligence
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
Pallamreddy Sai Harshavardhan, Naveen Kumar M, Maraneni Akhil, and Senthil Velan S
IEEE
In today's society, abnormalities in human lungs are fairly frequent, including illnesses such as tuberculosis, edema, pneumonia, and others. Convolutional Neural Network (CNN), Visual Geometry Group (VGG), Densely Connected Network (DenseNet), Residual Neural Network (ResNet) are a few image processing algorithms deployed for lung disease prediction. This study involves developing a multi-classification algorithm for predicting multiple lung diseases using X-ray images. The project employed several Python libraries such as TensorFlow, Keras, and NumPy. The deep learning frameworks used in the project were VGG-16, ResNet-50, and DenseNet-121, and the dataset used was the NIH chest radiographs dataset obtained from the Kaggle repository. The accuracy values of ResNet-50, VGG-16, and DenseNet-121 were found to be 56%, 75%, and 88%, respectively. Preprocessing procedures such as data augmentation, feature selection, and dimensionality reduction are crucial for accurate predictions from X-ray images. Overall, the study aimed to provide accurate and reliable predictions of lung diseases using X-ray images and demonstrated the efficacy of using the specified set of deep learning techniques.
Shreya Soni, Shruti Chaubey, Suchita Parira, and Senthil Velan S
IEEE
Facial expressions play an important role in conveying emotions, especially in human-machine interaction. Automatic facial expression recognition (FER) systems have numerous potential applications, such as detecting mental disorders, understanding human behavior, and generating synthetic human expressions. However, achieving high recognition rates remains a challenging task. In the literature, two popular approaches for automatic FER are based on geometry and appearance. The FER process typically consists of four stages, namely pre-processing, face detection, feature extraction, and expression classification. In our project, we utilized various deep learning techniques, specifically convolutional neural networks, to detect seven essential human emotions: anger, disgust, fear, happiness, sadness, surprise, and neutrality. Furthermore, our aim was to predict suicidal tendencies based on the detected emotions since depression is the primary cause of suicide. Detecting emotions in depressed individuals could facilitate their monitoring and help prevent suicide risk by forecasting the rate of suicidal intentions based on their emotional state.
Senthil Velan S, Preethika Reddy S, Preethi Bheemshankar Talwar, and Gomathi R D
IEEE
Predictive Machine Learning techniques provide mechanisms for the computing machines to analyze and understand the knowledge inherently embedded in the given dataset. This unique technique can be effectively used in understanding the quality of the question papers compiled for Continuous Assessment Tests (CATs). Understanding the quality of the question papers requires a systematic and software engineering-based approach. In this research work, a few of the CAT question papers have been carefully picked and used for applying the proposed methodology. Well, defined set of features, properties, and quality attributes have been applied in the proposed model and mapped into each other for providing a better understanding of the question paper's quality. Using metrics the analysis of quality for the given CAT paper is attempted. Based on the application of the model, it was able to clearly predict and understand the quality levels of the given set of question papers.
Divya Singh and Senthil Velan S
IEEE
Prediction or forecasting is the technique of uncovering the forth coming event by learning and obtaining experience through data collected from historical happenings and results. Prediction is used in almost every field today be it retail, healthcare, finance, marketing, travel, insurance, telecommunications, pharmaceuticals, language processing, and other fields. Analytics can be based on the collected data and is commonly and broadly used for analyzing and extracting knowledge obtained from data collected through social inter-networking. Social media contains abundant amount of multifaceted information allowing users to evolve into successful content creators. Henceforth, they also eventually become the web content distributors. So, an online game exists, since only a few features are becoming popular and the other remaining items are not so popular. Prediction of popularity will be highly significant in inter-networking dimensions considering the properties of caching and replication. In this paper, based on the surveys obtained about games’ popularity methods and features that have decent forecasting capacity are utilized to develop an algorithm using support vector classification to predict the popularity of the game.
Senthil Velan S, Rubini P, and Sivaranjani S
IEEE
Spiking Neural Networks (SNN) are biologically inspired networks working on the principle of communication triggered while crossing of threshold potentials. During the COVID-19 pandemic, immunity has been acquired by the population in a geographical location by infections and immunizations. The Waning Immunity Model (WII) has been used to apply the method of SNNs so that the results of the model provide a better way of understanding its effects. The dataset considered in this research is for a time period of six months during the years of 2021 and 2022 focusing on the geographical location of India. Based on the proposed new model, the spike in the WII index is clearly evident in the first half of the time period under consideration, This model will help the healthcare and governments officials to plan for the booster doses to be administered to the human population for reinvigorating the antibodies effectively fighting the COVID-19 virus.
Divya Singh, Senthil Velan S, and Vijayakumar K
IEEE
Connotative Relationship of sentences and paragraphs provide a much-needed method that can extract the sentiment, emotion and knowledge of a writer. Identifying this inherent relationship from the hidden layers of English paragraphs will help in understanding the overall sentiments and emotions of the paragraphs written by various authors. In this context the Natural Language Processing features and packages available in Python can provide a simple and unique solution for the extraction of inferred sentiments and emotions expressed by authors. In this research, an initial attempt has been made to extract the hidden sentiments of individual sentences and paragraphs from English text written by authors. Based on the work done it was found that even though the sentences of a paragraph were providing negative connotations the overall sentiment of the complete paragraph showed a positive sentiment and vice versa. Hence, it is inferred that analyzing and finding the overall sentiment of the paragraphs makes better understanding on the connotative meaning expressed by the authors in their textual writings.
Senthil Velan. S, Rubini. P, and Surbhi Choudhary
IEEE
Waning Immunity is an important and relevant concept during these days as the COVID-19 pandemic is expected to become endemic in the coming months. By definition, Waning Immunity is the loss of protective antibodies over time and hence necessitates booster shots at regular intervals of time. This quantitative study is on proposition of a model for computing a newly defined metric called Waning Immunity Index (WII). The model takes into account the three group of people namely, susceptible, infected and recovered individuals from the COVID-19 infections. The required data can be collected from the Kaggle repository that contains information on infections, recovery, vaccination and booster doses given on the human population while considering a geographical location. The proposed model and its implementation have thrown light on the spread, control and effect of COVID-19 virus. Results of the proposed model and the measurement can help health officials to seamlessly plan the duration of booster doses administered on vaccinated population. A sample data has been prepared for testing the model and the application of the proposed metrics. Based on the results, it is found that vulnerability of the Waning Immunity increases steeply at some duration and gradually steadies in time.
Velan S Senthil and Venkata Reddy Poluru
IEEE
Powerful and resourceful Image Processing techniques can be applied in the field of manufacturing engineering to understand a good number of quality attributes. In this research, the focus has been to apply the edge detection algorithms of Canny, Prewitt and Sobel to identify the quality of hand grinding and surface grinding done on standard size steel metals flats. The edge detected images are compared with the reference good grinded surface images to understand the similarity between them. Multiple samples taken from a laboratory environment are considered for the comparison. Based on the results obtained it is found that Canny edge detection function is able to find a good number of defects in the given set of samples. It is also found that the grinding done in each case is only around 25% perfect even if the Sobel algorithm is used for the surface edge detections.
Rohan Ramachandran, Senthil Velan S, and Daifa Imtiyaz Wadekar
IEEE
On December 2019, the COVID-19 deadly virus emerged and affected a large population in the world which led to the increase in the death rate of infected patients. Changes in dietary habits and way of living leading from the implementation of lockdown during this pandemic had been detrimental on the nutritional health of individuals. A statistical study was conducted to determine the effect of COVID-19 lockdown on dietary habits, food consumption, and weight in different countries. The primary aim of this study bas been to conduct a statistical research to measure the impact of the pandemic on dietary habits of the humans. Retrospective study was made by the data collected from a data repository Kaggle with attributes of food status, nutrients and calories. By statistically analyzing the data using RStudio, it was possible to infer on the changes in dietary habits between countries of similarities. Based on the statistical analysis, it was found that the consumption of fat varies depending on the availability of meat or vegetables during the COVID-19 pandemic leading to vulnerability of human population with weaker consumption-based immunity against the disease.
Aysha Musthak Ahamed, Senthil Velan S, and Daifa Imtiyaz Wadekar
IEEE
The Covid-19 pandemic that created a situation of public health emergency worldwide called for an urgent requirement of vaccine to prevent further spread of the disease. Successful vaccines started coming into view publicly since December 2020 with Pfizer-BioNTech vaccine being the first with a 95 percent efficacy. The primary purpose of the research is to conduct a statistical analysis of the effect of the covid-19 vaccines on herd immunity. Different countries that have used similar vaccines have been paired and analyzed based on the number of the new cases, recovered cases and the amount of vaccines taken. A comparative study on the herd immunity of the countries before and after the intake of the vaccination is made, to analyze the effect of the vaccines.
Jayanth Vadlapati, S Senthil Velan, and Ewin Varghese
IEEE
In this research paper, we are going to see the profound scientific use of computer technology applied in the fields of AI and Machine Learning primarily focused on Image Processing and Pattern recognition. Techniques such as ours are widely used to recognize real life objects including human faces etc. Thus, using such techniques, we can recognize a person from pictures. Using face recognition modules from python's huge collection of libraries, we are able to train the model to recognize people while wearing masks. Since when masks are worn, half of the facial features are lost, therefore developing a technique to recognize faces in such way is crucial. This specific technology of face detection is used in biometrics, video surveillance, etc. Therefore it's at utmost importance to increase the security as well as efficiency whilst making the recognition faster.
Joanita DSouza and Senthil Velan S.
IEEE
Exploratory Data Analysis (EDA) is a field of data analysis used to visually represent the knowledge embedded deep in the given data set. The technique is widely used to generate inferences from a given data set. Data set of current pandemic, the COVID-19 is widely made available by the standard dataset repository. EDA can be applied to these standard dataset to generate inferences. In this paper, data visualization technique is applied to the dataset and is used to formulate patterns for better insights on the effects of the pandemic with respect to the variables/ labels given in the dataset. A Web application tool called Jupyter Notebook is used to generate graphs using python language as it consists of libraries which are used for the process of EDA and the visualization is depicted for the attributes showing higher correlation. Based on the graphs obtained, we can draw conclusions from the current situation based on the data available, understand why a certain variable is increasing/decreasing with respect to another and what can be done to improve the drawbacks found.
Jumana Nagaria and Senthil Velan S
IEEE
In Exploratory Data Analysis (EDA) the given large data is visually analyzed to extract the embedded deep. Application of the technique has a wide range and aids in the informed decision making abilities of the managers. In an educational institution, the success of its imbibing model is usually measured using the career opportunities of the graduates. Hence, the placement data has an important relevance for the future plan and growth. Quite a good amount of information can be gained by all the stakeholder by carefully looking at this information. In this context, the technique of EDA can be used to visually analyze the placement of students in a higher educational institution. In this paper the data about the placement of student is visually analyzed to generate inferences using mathematical models. Based on the study it was found that student with MBA specialization in Mkt&Fin are highly placed, a vast majority of the students have Commerce and Management degrees. The score on the employability test don't seem to have a major impact on the placement of students.
S Senthil Velan and Venkata Reddy Poluru
IEEE
Digital Image Processing is an important field of computer science significantly focused on providing solutions to extract patterns and features from the given set of images. Image processing algorithms can very useful and can provide information such as edge detections in the given set of welded images. This technique to identify the edges can play a role in determining the quality of welding and thereby reducing the cost of maintenance. Edge detection algorithms such as Canny, Prewitt and Sobel are very efficient in identifying the spread of edges in the given greyscale welded images. In this research, the edge detections of the given set of greyscale welded images are compared with the edge detections of a quality welding images. The types of welding were considered for the comparison, namely, ARC and MIG welding of steel joints. Based on the results, it was found that samples obtained from the laboratory tests were have a quality of only 25% match in comparison with reference good welds.
Sabeel Ashfaq Khan and S. Senthil Velan
IEEE
Exploratory Data Analysis (EDA) is a data analysis technique that can be used to visually represent the knowledge embedded deep in the given data set. The application of this technique can be used in medical data processing for the betterment of the offered services of healthcare providers. In women, breast cancer has become prevalent and requires proper methods to identify the possibility of occurrence at an early stage for treatment and cure. In this context, visualization technique of EDA can be applied to the existing datasets for learning and prediction. In this research work, the analysis aims on finding predominant features which would be helpful in predicting whether the tumor is benign or malignant. To achieve this goal various graphical techniques are used for a better visual understanding. Based on the results obtained it can be found that EDA plays a significant role in describing the data using statistical and visualization analysis without making any speculations about the content. With this dataset and EDA approach we have achieved a clear understanding about the significant features needed to predict if the tumor is benign or malignant.
Senthil Velan S.
IEEE
Internet of Things (IoT) has become the buzzword for the development of Smart City and its applications. In this context, development of supporting software forms the core part of the IoT infrastructure. A Middleware sits in between the IoT devices and interacts between them to exchange data among the components of the automated architecture. The Middleware services include hand shaking, data transfer and security among its core set of functionalities. It also includes cross-cutting functional services such as authentication, logging and caching. A software that can run these Middleware services requires a careful choice of a good software modelling technique. Aspect-Oriented Programming (AOP) is a software development methodology that can be used to independently encapsulate the core and cross-cutting functionalities of the Middleware services of the IoT infrastructure. In this paper, an attempt has been made using a simulation environment to independently model the two orthogonal functionalities of the Middleware with the focus to improve its modularity. Further, a quantitative measurement of the core design property of cohesion has been done to infer on the improvement in the reusability of the modules encapsulated in the Middleware of IoT. Based on the measurement, it was found that the modularity and reusability of functionalities in the Middleware software has improved in the AspectJ version compared to its equivalent Java version.
Joanita Dsouza and Senthil Velan
IEEE
Preventive Maintenance is the new buzzword to upkeep an enterprise application in a real world scenario. It has also become a necessity to predict the faults that could occur between the different transfer nodes of the enterprise application. The features of preventive maintenance which includes methodical study, estimation, the notion of time, the diagnosis of faults and so on can be analyzed using well designed and structured methods of relevance to the application domain. In this scenario, the technologies which are focused are diagnosis of faults and activities, prediction of the states and monitoring the conditions. There is a large amount of data which is being used for preventive maintenance. The flow of packets between transfer nodes can be stored and this data can be used for purpose of training and testing. Further, machine learning algorithms can be implemented based on the features present in the historical data. The advantage on the study of preventive maintenance is that it can aid in reduction of cost which is the primary motive of each organization existing toady. By doing so the results can be applied to various organizations to resolve failures that could possibly occur in the future well in advance. This paper explains the application of machine learning algorithms for the detection of fault in transfer noted using preventive maintenance in an organization.
Laura Elezabeth and Senthil Velan S
IEEE
In any enterprise level organization, it is necessary to have an efficient and well-designed workflow for the seamless movement of task across its various participating units. In such a situation, the management of the workflow of the large scale enterprise is centralized on an integrated information system called Enterprise Resource Planning (ERP). It is incorporated across various functional areas of the company focussing on solving the needs and resources of the units in the organization. Such a kind of system must be able to manage data from all the departments in their large cloud storage. An implementation of such a scale and value has positive benefits to the overall performance of the large work flow based organization. In this paper we will see the implementation, drawbacks and how to improve the workflow management system in a way that is best suited for an extremely large scale organization.
S. Senthil Velan
IEEE
Aspect Oriented Programming (AOP) is a software development methodology that aims to improve the modularity by encapsulating the cross-cutting concerns into modular units called aspects. Inheritance of classes and aspects play a vital role while redefining the units of encapsulation. In this research, the impact of using multi-level inheritance in Aspect Oriented Soft-ware is quantitatively evaluated, with an extended and validated metric, namely the Weighted Average Depth of Inheritance. The metric has been applied on two equivalent OO and AO versions of a multi-level inheritance based Banking application case study which was developed using Java and AspectJ programming languages respectively. Based on the measurement it was found that the AspectJ version exhibits reduced cognitive complexity of computational intelligence compared to its equivalent Java version.
Senthil Velan S.
IEEE
The proponents of Aspect Oriented Software Development (AOSD) methodology have done a tremendous amount of work to bring out the positive effects of its adoption using quantitative assessment. A structured assessment of the methodology requires a well-defined quality model. In this paper, an AI agent based quality model has been proposed to evaluate the effect of aspectization. The model has been applied on already existing and equivalent versions of object oriented and aspect oriented case study application, university automation software. Specific metrics for the software design properties have been measured using AI agents for the different versions and were used to infer upon the effect on quality. Based on the initial measurement, it was found that aspectization has positively improved all the three quality attributes defined in the quality model. The attributes of evolution and reusability showed significant improvement in quality due to the effect of aspectization.
Velan S Senthil
IEEE
Aspect Oriented Software Development (AOSD) is a promising methodology for efficiently capturing the crosscutting functionalities (concerns) into modular units called aspects, thereby reducing maintainability, and increasing reusability and ease of software evolution. Since inheritance of classes and aspects plays a vital role in defining the units of encapsulation, it is essential that the impact of introducing inheritance in AOSD need to be quantitatively captured using core design level metrics and mapped onto higher level quality attributes. In this research work, a new set of metrics have been proposed to quantify the impact of using multi-level inheritance hierarchies in an Aspect Oriented software. A model that captures the reusability using inheritance of aspects has been used for defining and applying the proposed metrics and later relates them onto higher level quality attributes. Further, the proposed metrics and model have been applied to AspectJ versions of an SOA based case study application. Based on the measurement, it was inferred that aspectization has improved the higher level quality attributes of reusability, modularity and maintainability of the case study over its versions.
Senthil Velan S and Sam Jaffray M
Science Publishing Corporation
Service Oriented Architecture (SOA) facilitates seamless application integration through standards-based predefined web services. During integration, Business Process Execution Language (BPEL) plays a vital role in composing existing Web Services thereby achieving a service based workflow model. Due to frequently changing business requirements, it becomes very much essential for an SOA application to have the capability to dynamically bind with an alternate service rather than statically fixing the services in a given composition. However, BPEL lacks support for the run-time inclusion of a new Web Service or functionality. Aspects overcome this limitation by providing support to independently encapsulate the cross-cutting functionalities by separating them from the core business logic. Using AOP, it is possible to achieve dynamic binding in web service composition. To illustrate the embedding of AOP constructs into a BPEL process, this paper implements a case study on distributed e-HealthCare system. Further, two core design level properties namely, cohesion and coupling have been measured and the impact of introduction of AO into a composed BPEL process has also been discussed. Empirical evaluation of the design level properties shows that cohesion improves by the introduction of AOP in BPEL.
S. Senthil Velan and R.V. Sindhu Priya
IEEE
Aspect oriented programming is a technique that separates the core and cross-cutting concerns thereby increasing the modularity of the software. One of the important problem faced in AOP is the possibility for occurrence of interference between modeled artifacts. Interference occurs when the logic of one artifact interfere with the logic of another artifacts. Detecting and removing interferences in the analysis phase of software development will improve the design reflected by the desirable execution of software. This paper provides a new methodology to detect and resolve the interference in during analysis phase of Aspect oriented software development. A Java based tool called Early Aspect Oriented Analysis (EAOIA) has been developed to automate the identification and removal of interference. An existing AO analysis method namely, Theme approach is used to identify the base and the cross-cutting themes from the given set of requirements. The EAOIA tool identifies the existence of two types of interference by analyzing the intermediate code generated from the requirements. The cause and type of the interferences are identified and solution for its removal is suggested to the user. The user can then apply and verify the removal of identified interferences. The tool was able to successfully identify and remove the occurrence of data-flow and control-flow interferences in a case study application.
R Teebiga and S Senthil Velan
IEEE
Design pattern is a template for solving commonly occurring problems in similar situations. Design patterns can be easily implemented using Java programming language. To increase the software modularity design patterns can be used to model cross-cutting concerns in AspectJ programming. The effectiveness of using design patterns in both AspectJ and Java programming to encapsulate functional and nonfunctional design elements can provide insight onto the applicability of design patterns. Hence, a new methodology is proposed to compare the design pattern implementation of functional ad non-functional design elements as classes and aspects. An AODPE tool measured the design patterns implementation of functional and non-functional design elements. Based on the initial pointers of comparison it was found that AspectJ implementation shows better design properties such as size, inheritance and coupling while implementing cross-cutting functional and non-functional design elements as aspects.