Tamizh Ponni VP

IB Design Teacher, Oakridge International School

Abstract:
his paper is written bearing in mind the impact and rapid growth of Artificial Intelligence in educational technology. It talks about the intelligent systems that have taken over the data management in the field of education. From the enigmatic machines that cracked codes to the current systems that are trying to explore alternate realities and worlds, technology has created a massive revolution in all areas. As educators, it is an absolute necessity for us to be aware of the impact these advancements can have in Education. There has always been an emphasis on the 21st century learning skills-creativity, collaboration, critical thinking and communication to create future-ready citizens. When it comes to the ever-changing world of technology and computing, selecting the right tool that stands in sync with where Technology stands today, is important. Many AI concepts and tools are still in the labs and are yet to be explored by teachers and brought into the classrooms. Exploring the possibility of providing a personal learning set-up to both the teaching and learning community is the need of the hour. Technology can no longer be treated in isolation but as an integrated operative tool to make students, life-long learners. There are a number of ways invented and experimented to control technology by human intelligence. Today with virtual classrooms, we have reached a phase that cannot be oblivious to technology and schools have started considering the need to have a strong technical infrastructure to run their operations-both admin and academic irrespective of space and time. This paper talks about effectively identifying the appropriate AI tool and designing and implementing the right teaching methodologies with the right tools and supporting technologies that are very much going to be deciding factors between success and failure of our objectives in preparing our children for new-age professions.


Keywords - Artificial InteIIigence; Machine Learning; Learning ModeIs;Design Thinking; VirtuaI ReaIity, Augmented Reality;

Mousoomi Bora¹ and Rupam Baruah²

1. Department of Computer Science and Engineering, Assam Kaziranga University, Jorhat, Assam, India, mousoomi@kazirangauniversity.in
2. Department of Computer Science and Engineering, Jorhat Engineering College, Jorhat, Assam, India, rupam.baruah.jec@gmail.com

Abstract:
Artificial Intelligence and Machine Learning are upfront technologies to predict breast cancer in its initial phase. The pathological image and numerical data are used to train the Machine Learning models so that the breast tumors can be detected in its premature stage. The objective of this research is to predict Breast cancer in its early stage by means of two famous machine learning algorithms viz. ‘K Means’ and ‘Support Vector Machine (SVM)’. The dataset used in this study has been treated with a dimensionality reduction technique called ‘Principal Component Analysis (PCA)’ as a part of pre-processing step. The ‘Elbow method’ is used with ‘KMeans’ Clustering to compute the cluster quality. The experimental results demonstrate that the ‘Support Vector Machine’ (SVM) give superior performance than the ‘KMeans algorithm’.


Keywords - Principal Component Analysis (PCA), Elbow method, KMeans and Support Vector Machine

Ananya Das*ϯ · Subhashis Chatterjeeϯ

ϯDepartment of Mathematics and Computing, Indian Institute of Technology (Indian School of Mines), Dhanbad, Jharkhand, India.
*Corresponding author, Email: ananyadas16@gmail.com

Abstract:
Classification of cancer, specifically based on gene expression pro-files, is known to be the guiding light for addressing the pivotal issues related to drug discovery and cancer diagnosis. It helps to provide insights on the causes of cancer and its treatment. Gene expression data is often represented by huge amount of gene data with a limited quantity of samples. Another challenge is to identify the group of genes that are correlated and are highly expressed in dif-ferent type of tumor cells. Prompted by the above challenges, a hybrid method of clustering and classification of gene expression RNA sequence data into dif-ferent cancer types using machine learning techniques is proposed in this work. Machine learning techniques have their own properties viz. ability to find the vital cancer genes for classification, showcasing the correlation among the genes and classifying it into different cancer types. Owing to these properties, the current research article is categorized into four distinct modules: Pre-processing, Feature Selection, Gene Clustering and Classification using a re-duced set of feature vectors. In this work, Recursive feature elimination serve as an efficient tool to discover the significant genes. Clustering is an important and promising tool for analyzing gene expression data. Subsequently, a gene simi-larity matrix is computed based on Mahalanobis distance measure which serves as the input for the self-organizing map for the clustering of the data. Using novel concept of correlation within the cluster, a reduced set of gene vectors is obtained which further serve as the input for the neural network classifier. Fi-nally, the output classifies the data into four different cancer types. Comparison of performance is made between the proposed work and other machine learning algorithms. The result demonstrates that the hybrid technique of classifying the gene expression data is an efficient approach in terms of accuracy thereby out-performing other machine learning models.


Keywords - Gene Expression Data, Recursive Feature Elimination, Mahalanobis distance, Clustering, Classification

Devashree A. Joshi

Pune, Maharashtra, India, devashree31198@gmail.com

Abstract:
During the COVID-19 pandemic, the circulation of misinformation and fake news has generated a lot of fact discrepancies and scientific oversights. Our research aims to comprehensively assess the spread of misinformation regarding COVID-19 and analyse its reach across demographic parameters like age group, gender and country of residence. The analytics was performed using various open-source technologies like Python, Tableau, R Studio by generating diverse visual plots and Word Clouds. For experimental purposes, we considered India and USA as countries of focus and the data was collected accordingly. Furthermore, an original Survey was designed and conducted to trace the reach of viral, verbatim misinformation articles in both the countries. We studied the misinformation data across parameters like – types of misinformation, motives, to name a few. Our research proved to be of practical relevance and it is gauged to be beneficial to strategize mitigation measures of the misinformation infodemic prevalent during a devastating pandemic.


Keywords - mis information analytics, COVID-19 misinformation, fake-news analytics, disinformation, survey analytics, big-data analytics

Sachin Kulkarni1[0000-0002-4227-5390] and Kishor Kolhe1[0000-0003-0205-4814]

Dr. Vishwanath Karad MIT World Peace University, Pune, India, lncs@springer.com


Abstract:
The Blockchain technology is one of the recent trends to have impact on many technological advancements in the fields of banking and finance, digital currency, security, intellectual property, and many more. The academia and com-mercial industries are the important partners of block-chain technology to get the benefit and grow exponentially. The education systems are evolving at the global stage with blended learning model and the distributed nature of blockchain tech-nique certainly helps achieve integrity whilst operating on the data and infor-mation. The recent scenarios of pandemic have boosted the education systems to move towards digital engagement and blockchain will have a major role to play. This paper puts forth the survey of research work carried out in the field of edu-cation and the impact it has had by the inclusion of blockchain technology. The progressive research study in this paper deals with the questions arising on the implementation blockchain involvement in education sector by going beyond the certificate generation.

Keywords - Blockchain in education, distributed systems, educational technol-ogy, consensus protocols, certification

Badari Narayan V S1, Dr.Siddappa M2

Research Scholar, Department of Computer Science, SSIT, SSAHE, Tumakuru
Professor and Head, Department of Computer Science, SSIT, SSAHE, Tumakuru.


Abstract:
Cloud computing is a game changer in the field of data storage service, computer application service, Data provisioning service and many more. More and more organizations are moving towards cloud for different types of services requirement. Cloud computing is an Internet-based computer technology that refers to the use of Computing resources i.e. hardware and software, available on demand as pay and use service over the Internet. Data on the cloud server is the real concern for client as he loses control over the data. Here we highlight the client security concern with respect to data. Overall Cloud functionality has been divided into four layers. Each layer is responsible to carry out specific activity. At each layer data will be handled by different person and at different location. At each layer we propose a four layered security for data. The first layer is cloud user layer, in that a robust and dynamic method has been proposed and implemented for access control. Access control at user layer (client side) to mitigate internal threat is implemented using mathematical model and decision tree.


Keywords - cloud data, Security, Access control, DMAC, ANOVA, Decision tree.

Naveed ul Islam¹ and R. Sunitha²

1. Department of Computer Science, School of Engineering and Technology, Pondicherry University, Pondicherry 605014, India, nav33d.cs@gmail.com
2. Department of Computer Science, School of Engineering and Technology, Pondicherry University, Pondicherry 605014, India, sunitha.pondiuni@gmail.com

Abstract:
Real world systems are often represented as graphs (networks) of entities with relationships and features shared between them. Social Network Analysis (SNA), Machine Learning and Graph Analytics methods are used to extract latent insights and knowledge from such graphs for downstream applications. Community Detection is an SNA approach to uncover the structural and functional organization of such graphs. Owing to the power of deep learning, a variety of approaches under Graph Representation Learning have been proposed that exploit the graph data to learn low-dimensional yet high-quality representations that generalize for subsequent tasks such as node classification, link prediction and clustering. This work proposes a Gaussian-Mixture based graph representation model using variational inference to the task of community detection. The proposed model encourages an improved low-dimensional learning and community-friendly embedding through two separate mappings from the original graph space: 1) from a latent representation to reconstruction space and 2) from the latent representation to a community-friendly space. Such a separation is conceptually theorized to yield neat communities and also overcome the limitations of previous approaches proposed for the task of community detection. The advantages of the model are a better representation of the underlying data distribution, a variational inference scheme to confirm the a priori assumptions and an architecture-independent learning methodology specific to the task of community detection.


Keywords - Social Network Analysis, Graph Neural Network, Graph Representation Learning, Community Detection, Deep Learning, Variational Inference.

N. Kiran Venkatesh and Dr. Geetha Ganapathi

Department of Applied Maths & Computational Sciences, P S G College of Technology, Coimbatore 641 004, India

Abstract:
Most of the deaths are caused due to road tragedy.Amidst various types of road accidents, bike accidents are usual and cause serious injuries. Main protection for biker’s is wearing helmet. Road tragedy happens in many countries because biker’s didn’t wear helmet and fail to follow the rules.Detection of helmet defaulters is a highly required to guarantee safety precautions, but stretching jobs caused by different problems such as poor & low quality of phone video, phone movement while taking video,glaring effects,differing climatic environment, traffic jam etc. Computer Vision & Deep Learning algorithms like CNN is used for detecting bikers who are breaking helmet laws is presented. The structure consists of two stages namely bike detection and helmet vs. no-helmet classification. A structure for helmet detection of bikers driving without helmets in Mobile Phone videos is presented. In the structure, initially Haar Cascade Classifier is used to choose bikers as a support among the moving vehicles. Then CNN is used to select Bikes. Again, CNN is used on top quarter part for further identification of biker’s driving without a helmet. The production of the structure is evaluated on 5 videos. Structure is decided based on accuracy. The experiments on mobile videos produced 82.7% accuracy with data shuffling done and auc_roc_score as 0.42% on an average without data shuffling and thus demonstrated the efficiency of the structure with very limited and low quality data.


Keywords - helmet detection, OpenCV, Haar Cascade, Image Data Augmentation, CNN, Supervised Learning, Keras, Checkpoints.

Neetu Singh, Ashwini Naik and Ayush Saxena

Department of Electronics & Telecomm, R.A.I.T,Nerul, Navi Mumbai, India, singhneetu07@gmail.com,
Department of Electronics & Telecomm, R.A.I.T,Nerul,Navi Mumbai,India, ashwini.parwatikar @gmail.com
Department of Electronics & Telecomm, R.A.I.T. Nerul,Navi Mumbai, India, ayushsaxena3108@gmail.com

Abstract:
In this design ultrathin and wideband absorberbased on two-layered frequency selective surface (FSS) is presented. The structure consists of unit cell printed on two cascaded layers. Each layer has four metal strips bend at ninety degree intertwined with each other embedded with chip resistor. These bend metal strips loaded with chip resistor are imprinted on the substate, which are to impart resistive loss. Both the layer has identical geometry but different dimensions. The substrate material selected is FR4, which have a dielectric constant of 4.4 and dielectric loss tangent is 0.02. The thickness of two cascaded layers are respectively 2mm and 2.4mm. The designed structure is symmetrical in plane and therefore it becomes polarization insensitive for TE-TM polarized incident wave. The full wave simulations represent that the designed structure generates 3 poles, that are caused by the upper FSS layer, the upper FSS and the lower FSS coupling, and the lower FSS layer, respectively. These three poles widened the absorption band. The proposed multilayer absorber provides reduction in reflection coefficient of 10 dB with broad bandwidth of 6.46 GHz in the frequency range of 11.54 - 18.00 GHz. Additionally, the width of the structure is 4.4 mm which is very thin, only 0.16 of wavelength at lowest frequency. The developed broadband and thin absorber provides an feasible and effective solution for stealth applications.


Keywords - FSS, Ultrathin, Wideband, Absorber, Polarization insensitive.

Ishi Khamesra, Srinivasarao

Valluru Hyderabad, India, Hyderabad, India

Abstract:
We live in an age of accelerated Digital Transformation spanning across diversified industries barring none. The telecom industry in particular has witnessed a tectonic shift in enabling the digitalization of services over the past decade, such digital service culture can always bring fraudulent practices to the fore exposing customers to be victims of fraud. Safeguarding the customer information and preventing fraud is of paramount importance across all. The advent of cutting edge Artificial Intelligence and Natural Language Processing solutions mark the dawn of a new era in predicting and preventing Telecom device fraud, “the focus of this paper”. Often risk score thresholds lead to an incomprehensive way of fraudulent orders detection, service providers in such scenarios delve into additional customer attributes to ascertain order legitimacy. The investigation process calls for drafting summary remarks for each such order investigation. Service providers can leverage this data that benefits them in multiple ways (a) to optimize and reduce investigation cycle time and (b) to take corrective actions through decision score based investigative methods.


Keywords - Cosine Similarity, Extreme Gradient Boosting (XGBoost), Grid Search CV, Random Forest, Recursive Feature Elimination (RFE), Countvectorizer, VADER, Stemming, Lemmatization

Shubham Agrawal, Aditi Chikkali, Rakhi Akul, Gayatri Degaonkar, Prof. Rashmi Dixit

Dept. of Computer Science Walchand Institute of Technology Solapur, India, shubhamvagrawal@gmail.com
Dept. of Computer Science Walchand Institute of Technology Solapur, India, aditichikkali45@gmail.com
Dept. of Computer Science Walchand Institute of Technology Solapur, India, akulrakhi@gmail.com
Dept. of Computer Science Walchand Institute of Technology Solapur, India, gayatridegaonkar77@gmail.com
Dept. of Computer Science Walchand Institute of Technology Solapur, India, rkdixit@witsolapur.org

Abstract:
Traditionally CCTV cameras are used only for theft prevention and security purpose; however, the video footage can be used for various kinds of information mining purposes using video analytics and artificial intelligence. Our solution is an Artificial Intelligence-driven solution that can track objects via CCTV Footage. We can track any object that the business would require to incorporate intelligence in business and thereby help in strategy building and making informed decisions. This solution helps in identifying and tracking objects that we need to keep track of in a business such as the places like warehouses, manufacturing industries, retail stores and even some kind of secure areas like offices that process confidential information. People or vehicles entering such secure areas can be tracked and restricted entry accordingly. A video analytics system can generate heat maps of a store to point out what proportion of time a customer dwells at a specific product, thus establishing its demand. By understanding the demography of the audience, and by observing the area in which they tend to concentrate more, the store owners can make changes in the inventory strategically. The system can also determine whether a store or a facility is understaffed, and can raise an alarm to call in more employees. The solution can also be used for security purposes, for example, tracking the movement of consignments, checking on unauthorized personnel in some restricted area by tracking the faces as well as vehicles going inside. The same technology offers license plate recognition, that can be employed by corporate offices, residential societies etc. The businesses employing these improvised techniques can not only track the number of customers, but also their shopping patterns, behavioral statistics, and a lot more.


Keywords - Object Detection, OpenCV, Python, Computer Vision, Video Analytics, Artificial Intelligence

Dr. Vipul Aggarwal, Abhishek Bhardwaj, Harbani Gill, Parul Naib

harbanis@hotmai.com

Abstract:
The analysis of images for forensic research is being done increasingly to detect any outliers and trends which are not possible through structured data and feature analysis. Image analytics leverages cognitive analytics, advanced artificial intelligence and machine learning techniques in order to analyse massive amount of visual unstructured data that is generated. The current paper discusses how AB-PMJAY, the world’s largest state funded health insurance scheme, leverages image analytics for fraud prevention and control. The paper looks at the image analytics algorithms being leveraged by the National Health Authority to analyse health claims’ image data under PM-JAY and how they contribute to strengthening the overall fraud detection and control strategy at NHA.


Keywords - Fraud detection, Visual Data, Image Data, Radiological, Image analytics, Compression Function