Mahesh Manchanda
@gehu.ac.in
Professor, CSE
Graphic Era Hill University
RESEARCH, TEACHING, or OTHER INTERESTS
Artificial Intelligence, Computer Science Applications, Computer Science, Computer Engineering
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
Supriya, Ashutosh Shukla, and Mahesh Manchanda
IEEE
Agriculture is the primary source of assistance for all the key sections of our country, INDIA. Population growth and Urbanization are the two major factors that create serious challenges for this field. According to the report of the Food and Agriculture Organization of the United Nations, by the next two decades, there will be a need for IT infrastructure to handle the rapidly expanding dietary needs of the two billion human population. Precision agriculture is the smart aid to resolving these issues. it allows for more efficient use of resources. It helps farmers to meet the growing demand for food sustainably by implementing smart IT techniques for several activities like Site-specific planting and fertilization, Crop monitoring, pest management, and variable rate irrigation. Autonomous vehicles, etc. keeping this in view a comparative study on advanced machine learning. Support vector machine, K-Nearest Neighbor, Artificial Neural Network, and Decision Tree algorithms have been implemented on MATLAB TOOL for early-stage identification of powdery mildew on cherry leaf. Hereafter considering various performance metrics like Confusion metrics parameters (accuracy, recall, precision), and ROC & AUC curve, it is found that ANN provides the best results of 91% on 1,200 images of the standard open source repository.
Himani Saraswat, Mahesh Manchanda, and Sanjay Jasola
Institution of Engineering and Technology (IET)
AbstractPredictive demand forecasting plays a pivotal role in optimizing supply chain management, enabling businesses to effectively allocate resources and minimize operational inefficiencies. This paper introduces a novel approach to enhancing demand forecasting processes by leveraging the Ethereum virtual machine within a blockchain framework. The proposed system capitalizes on the inherent security, transparency, and decentralized nature of blockchain technology to create a secure and efficient platform for predictive demand forecasting. The system leverages the Ethereum virtual machine to establish a secure, decentralized, and tamper‐resistant platform for demand prediction while ensuring data integrity and privacy. By utilizing the capabilities of smart contracts and decentralized applications within the Ethereum ecosystem, the proposed system offers an efficient and transparent solution for demand forecasting challenges. The current research focused on Ethereum virtual machine characteristics, features, components, and implementation details. A secured framework for the prediction of demand forecasting systems is proposed. Finally, the authors tried to validate and optimize the gas cost by using distinguished statistics and analysis.
Supriya, Ashutosh Shukla, and Mahesh Manchanda
IEEE
Inertial sensors embedded within smart devices are emerging research propel that gather information about people's actions. This valuable information can be further used to predict and analyses human behaviors using the latest techniques like machine learning and deep learning. The predictions made through these models have great importance in the field of medical medicine, like caring about elderly and mentally retarded patients and, they can also be used in security surveillance applications. In this proposed approach, a daily living Activity Recognition dataset, MHEALTH from Kaggle repository, which is built from the recordings of ten volunteers through the use of four inertial sensors of mobile devices placed at various body positions of volunteers. A hybrid CNN-LSTM deep learning model is implemented using Python libraries such as keras, pandas, tensorflow, numpy, etc. this model is a layered architecture of Convolutional Neural Network (CNN) to extract prime features from input data and to predict human activity in the long short-term memory (LSTM). The experimental results of the proposed model after hypertunning achieved accuracy of 98% over the MHEALTH dataset
Jyoti Agarwal, Shelly Gupta, Neha Sharma, and Mahesh Manchanda
IEEE
Plant diseases are emerging problem in agriculture sector which occurs due to various bacteria, viruses etc. There is lack of awareness among the farmers about these kinds of diseases which has adverse effect on crop production rate. One of the very important crops that plays significant role in Indian economy is tomato. In India, production of tomato crop is vast which has direct impact on Indian economy. Various kind of diseases can be cultivated in tomato plants also which can be harmful for the crop and farmers face economical loses. Due to these issues, it is important to detect tomato leaf diseases to prevent the crop as well as economic damages. The purpose of this study is to suggest an easy and precise method to identify and categorize diseases of tomato leaves. For this reason, a CNN method is applied as they employ automatic feature extraction as well as classification of the input image into different classes of diseases. Experiment is done on an online dataset in which images are classified into 10 different types of diseases. Proposed model was able to receive 93% of accurate results for detecting the correct diseases in tomato leaves which shows that proposed CNN model can be used as a feasible and efficient technique for identifying tomato leaf diseases in diverse circumstances.
Gagan Dangwal, Priya Matta, Sudhanshu Maurya, Sanjeev Kukreti, and Mahesh Manchanda
IEEE
This research paper presents the development and execution of a novel home automation system operated through voice commands using Google Assistant. The system allows users to control home appliances, such as lights and fans, from anywhere using their smartphones or other smart devices. By connecting the home appliances to the internet, they can be easily accessed and monitored in real-time, making it an ideal solution for Internet of Things (IoT) facilitated home automation. The proposed system adds an extra layer of security by integrating speaker identification and verification technologies, ensuring that only authorized users can give commands to the system. The paper presents the results of the implementation and testing of the system in a real-world home automation environment. The proposed system is flexible, low cost, and offers a high level of security, making it a potential solution for securing other IoT applications. The primary objective of this project is to demonstrate the potential of voice-controlled home automation systems for improving living conditions and the role of speaker recognition technologies in ensuring security.
Himani Saraswat, Mahesh Manchanda, and Sanjay Jasola
Elsevier BV
Deepak Banerjee, Vinay Kukreja, Mahesh Manchanda, Siddhant Thapliyal, and Shanmugasundaram Hariharan
IEEE
The abstract provides a thorough summary of a study that examines the effectiveness of various grapevine-related classes using a classification methodology. The model's capacity to precisely forecast positive cases inside classes is demonstrated by the research's precision scores, which range from 68.25% to 77.94%. Recall scores also range from 69.49% to 79.71%, demonstrating the model's skill in identifying genuine positive cases. The range of the harmonised F1-Score, which measures the harmony of recall and precision, is 69.35% to 78.01%. Model evaluations have been solidly grounded in large instances in each class, which vary from 590 to 750. The model's proficiency across all classes is demonstrated by the weighted average accuracy, recall, and F1-Score, each of which is impressively high at 85.40%, 86.00%, as well as 84.90%, respectively. Notably, the model's thorough understanding is demonstrated by the micro average accuracy, remembrance, and F1-Score, which reaches 73.78%. These findings demonstrate how well the model performs in accurately classifying a range of grapevine-related problems, significantly advancing tactics for managing and assessing grapevine health. The study's conclusions could have an impact on viticulture and farming strategies.
Varun Jindal, Vinay Kukreja, Shiva Mehta, Mahesh Manchanda, and Siddhant Thapliyal
IEEE
Early detection of leaf diseases is crucial in agricultural research to maximise production and maintain crop health. With the use of Convolutional Neural Networks (CNN) and federated learning across five different clients, each of which defines five severity levels of the illness, the current research reveals a ground-breaking method for comprehending Okra leaf disorders. Each client dataset provided a different perspective on the disease’s symptoms, which ranged in severity from moderate (1–25%) to severe (76–100%). Our study is drawn to the decentralised machine learning paradigm. It allows each client to analyse its data locally without sending the raw data to a centralised server, respecting data privacy principles. Insights from local models are included in the global model, which is realised via federated averaging to provide a comprehensive, unified viewpoint on Okra leaf disease detection. Initial findings show potential. The ranges for the customers’ macro, weighted, and micro averages were 87.21% to 93.96%, 88.27% to 93.94%, and 88.28% to 93.94%, respectively. Cltt-4 and Cltt-5 showed outstanding detection metrics, highlighting the model’s resilience. The federated averaging approach demonstrated its effectiveness by combining customer insights to create a comprehensive model with improved accuracy and recall rate, making it a priceless tool for farmers and agronomists. This study combines CNN’s strength with federated learning’s adaptability to provide a fresh perspective on Okra leaf diseases. It opens the door to a time of well-informed, decentralised, and privacy-protecting agricultural treatments that might completely alter how conditions are managed.
Amit Gupta, M. C. Lohani, and Mahesh Manchanda
Taru Publications
In the current scenario of digital world, every year the financial institutions have to face billions of dollars losses due to fraudulent transactions. Out of different categories of financial frauds credit card transaction fraud is the most common. To reduce the effect of these fraud transactions there is a need for a well-designed and secured fraud detection system with a state of art fraud detection model. Our work’s primary contribution is the creation of a fraud detection system that makes use of some mathematical usage of creating heat maps which is then enhanced with the use of a deep learning architecture and a sophisticated feature engineering method based on HCNN- Heat Map Convolutional Neural Network. HCNN is a model which create the heat maps for the imbalance data set without replicating the minor class records and without discarding major class records. The experimental findings show that our suggested technique is a practical and successful mechanism for detecting credit card fraud. The main objective of our model is to develop such a technique that can be used to detect and correctly classify more numbers of fraud transactions thus, our suggested technique, may detect considerably more fraudulent transactions than the benchmark methods with the accuracy of 91.7%.
Himani Saraswat, Sanjay Jasola, and Mahesh Manchanda
Frontier Scientific Publishing Pte Ltd
<p>Big data has become more and more popular, piquing the interest of both researchers and technologists as well as business executives. Despite its advantages, big data has a number of problems that necessitate a one-stop shop. Blockchain technology has seen a considerable increase in usage, which has had a substantial impact on its applications and led to a variety of useful outcomes. In the areas of identity, trust, decentralization, data-driven decisions, data ownerships, etc., there are notable game-changers. As a result, Blockchain is frequently acknowledged as an effective fix for big data issues. Among the solutions it suggests are decentralized private data management and digital property resolution. Together, two these technologies can develop beneficial solutions.<em></em></p>
Rahul Chauhan, Ayush Negi, and Mahesh Manchanda
IEEE
The goal of web scraping or data scraping is not only to search for or extract data from websites, but also to extract data in a systematic manner and make better decisions. Choose wisely and more. This can be one of the most difficult problems for users to overcome, as the internet is the largest source of knowledge, yet information found on the web cannot be used directly for data analysis and other processes because the language is not necessary. In order to get. It is vital information in a short period of time. Web scraping employs the usage of a machine, such as a computer, to browse a web page, extract data from the web page, and then store the extracted data. It is critical to exercise caution and ensure your safety.
Chandradeep Bhatt, Ritabh Sharma, Rahul Chauhan, Ashish Vishvakarma, Mahesh Manchanda, and Sanjay Sharma
IEEE
Learning computer science is tough for many students, especially the ones who don't have any knowledge about it previously. Just like other subjects, there are a lot of things that needs to be learned in Computer Science. It is often to seen that students find it difficult to cope up with the extreme change in their learning environment. There are many reasons why students struggle in this field. ‘Algorithms’ is one such topic that students find the most confusing. Lecturers on the other hand, also face problems in describing the algorithms to students. Based on several researches, Algorithm Visualization has been to have a positive impact on learning abilities of student. It is already known that a person can learn much more efficiently if it sees what is going on and how is it done. Pathfinding refers to the exploration of an optimal and shortest route, commencing from a designated initial point and concluding at a specified destination. Within the realm of artificial intelligence, pathfinding algorithms are specifically crafted to navigate through graph structures. These algorithms find utility in diverse domains like navigation systems, computer gaming, network architectures, and beyond. Their performance can be degraded by numerous factors such as the length of the path, size of the problem, obstacles and its distribution, data structures used and much more. Whenever a new path-finding algorithm is suggested, its performance is often investigated on the basis of the observations made when that algorithm is used. Proper design and analysis play an important role in giving an explanatory and precise evaluation. By conducting analysis of multiple research works, this study systematically categorized them based on their provisional designs, methodologies, and logical approaches. Throughout this process, this work discovered minor imperfections in these algorithms, which were frequently observed as well.
Supriya, Ashutosh Shukla, and Mahesh Manchanda
IEEE
Precision agriculture is a cutting technology in the field for agriculture, which deals with the challenges of the traditional methodology. This research work is a review of the recent studies published and discussed for detection of plant disease using ML & DL models on various plants dataset. This literature analysis is performed for publications from 2017 to 2022. More than 30 publications were selected and studied. In this present work, some of the existing ML & DL algorithms that are used to process the images for detecting crop diseases are discussed. The study highlights the results of the investigation of several existing ML and DL models, datasets used and gaps in work. Finally, this identified gaps that may decide the future direction of the research in this area. The purpose of this study is to provide knowledge for future research in building an accurate and effective classification plant diseases.
Umang Garg, Kartikey Jadli, Rahul Singh Pundir, Mahesh Manchanda, and Neha Gupta
IEEE
"Apple" proudly reflects Kashmir's fruit sector, which accounts for around 88 percent of the valley's overall fruit production. It forms the foundation of the Shopian district's economy in the Kashmir valley. Several factors, including the illness spread across the crops, have sadly prevented a considerable expansion in this industry's production over the past few years. India is an agricultural nation, and it is an important issue for the prompt detection of diseases among plants. Apple growers will be able to take the necessary actions or safeguards to protect the fruits from contamination if apple trees are discovered early. It takes a lot of time and effort to predict apple plant diseases using manual or traditional methods, which also requires providing lab diagnoses. But this is where the new technology comes into play; because of the development of machine learning and deep learning, it is now feasible to rapidly and accurately assess whether a plant is infected or not.
Umang Garg, Urmila, Rahul Singh Pundir, Mahesh Manchanda, and Neha Gupta
IEEE
Hepatitis B and Non-Alcoholic Fatty Liver Disease (NAFLD) are the most frequent and well-known liver disorders worldwide. Over time, these disorders cause liver cirrhosis, which finally leads to failure and major consequences, which can be fatal. More than a third of the world’s population has Hepatitis B and NAFLD. Blood tests are utilized to identify hepatitis B, and while the liver biopsy is the gold standard in diagnosing NAFLD, ultrasound scan pictures are employed in this work. The proposed method efficiently identifies or predicts Hepatitis B and NAFLD using the findings of this study, namely the application of machine learning and deep learning neural networks. Keyword: Fatty liver, Hepatitis B, Machine Learning, NAFLD.
Aditya Harbola, Mahesh Manchanda, and Deepti Negi
IEEE
In the information age with the reckless growth in the field of artificial intelligence and knowledge discovery a lot of research work is directed to solve challenges which were not at all thought in the assessment of computer science research area. Among many such problems, one such problem in the current scenario is of misinformation spread control and detection. A combined approach of natural language processing and Machine learning models can classify and detect misinformation. In this research paper we have used LSTM and BERT models to classify misinformation on a dataset and compared their accuracy and performance.
Umang Garg, Vineet Kukreti, Rahul Singh Pundir, Mahesh Manchanda, and Neha Gupta
IEEE
Forest fire is one of critical challenge for the survival of human, animal, and environment. Turkey is a particularly dangerous location for forest fires. Over the past two decades, there have been over 2,000 forest fires annually, each of which has caused at least one fatality.48% of them are due to the actions of people. When the rates of fires with unknown causes are factored in, this number jumps up to 71%. Based on the satellite data, this investigation has been carried out in turkey, comprising 39,289 forest fire data rows and 13 columns. This exhaustive study, which is based on the data points spanning the years 2000–2020 and seeks to predict forest fires, is the result of extensive research. The method of machine learning known as “Random Forest” serves as the foundation for the predictive model. According to the results of the experiments, the Random Forest Regressor performs better than other models in terms of accuracy rate, MAE, and RMSE value in pairs of 89.450/0,3.42460,6.198039 subsequently. This study will be of tremendous value to the Turkish government as well as the citizens, and it is expected to have an impact on the following permitting individuals to take safeguards against possible forest fires and get themselves ready for them should they occur. However, there is still an opportunity for development and improvisation in the situation with information.
Shruti Bhatla, Vikas Tripathi, Mahesh Manchanda, and Richa Gupta
AIP Publishing
The Covid-19 pandemic has turned out to be a global health crisis that has a deep impact on the outlook of how we understand our day to day life. Apart from the uncommon human cost, it has set off a profound impact on various sectors such as economy, education, etc. Due to this crisis a remarkable rise in the utilization of digital services has been experienced. Online learning, which became a panacea for this situation, is progressively observed as another standard in education. Though implementation of e-learning is promoted and is considered as a perfect solution too but still it put an additional strain on these infrastructures since all of these activities are carried out within the network, which leads to various issues such as network congestion, access delay, overloading, etc. This paper describes the various challenges which came in frame with the increased use of online learning in terms of networks during this global health crisis. © 2022 American Institute of Physics Inc.. All rights reserved.
Akshara Pande, Mahesh Manchanda, Hans Raj Bhat, Partha Sarathi Bairy, Navin Kumar, and Prashant Gahtori
Informa UK Limited
Abstract A phytoalexin, Resveratrol remains a legendary anticancer drug candidate in the archives of scientific literature. Although earlier wet-lab experiments rendering its multiple biological targets, for example, epidermal growth factors, Pro-apoptotic protein p53, sirtuins, and first apoptosis signal (Fas) receptor, Mouse double minute 2 (MDM2) ubiquitin-protein ligase, Estrogen receptor, Quinone reductase, etc. However, notwithstanding some notable successes, identification of an appropriate Resveratrol target(s) has remained a major challenge using physical methods, and hereby limiting its translation into an effective therapeutic(s). Thus, computational insights are much needed to establish proof-of-concept towards potential Resveratrol target(s) with minimum error rate, narrow down the search space, and to assess a more accurate Resveratrol signaling pathway/mechanism at the starting point. Herein, a brute-force technique combining computational receptor-, ligand-based virtual screening, and classification-based machine learning, reveals the precise mechanism of Resveratrol action. Overall, MDM2 ubiquitin-protein ligase (4OGN.pdb) and co-crystallized quinone reductases 2 (4QOH.pdb) were found two suitable drug targets in the case of Resveratrol derivatives. Indeed, carotenoid cleaving oxygenase together with later twos gave gigantic momentum in guiding the rational drug design of Resveratrol derivatives. These molecular modeling insights would be useful for Resveratrol lead optimization into a more precise science. Communicated by Ramaswamy H. Sarma
Aditya Harbola, Deepti Negi, Mahesh Manchanda, and Rajesh Kumar Kesharwani
Elsevier
Amit Gupta, M. C. Lohani, and Mahesh Manchanda
Taru Publications
Abstract This is the era, where the plastic money concept is widely adapted all over the world, but every new technology has its own loopholes also. In this scenario many types of anomalies can happen which can harm the user economically. These anomalies can be defined as frauds in financial sector. To detect these types of frauds, many techniques and models are proposed by the researchers. In this study the proposed work tries to implement an automated model using different machine learning techniques for the detection of these kinds of frauds, especially related to credit cards transactions. The proposed model applied four algorithms used in machine learning, namely Naive Bayes, Random Forest, Logistic Regression and SVM on a very large dataset to predict the fraud. Naive Bayes algorithm performance is outstanding for detection of credit card fraud among all the ML algorithms with the accuracy 80.4% and the area under the curve is 96.3%