Browse Items (2150 total)
Sort by:
-
Intricate Plane of Adversarial Attacks in Sustainable Territory and the Perils faced Machine Intelligent Models
The issue of model security and reliability in Artificial Intelligence (AI) is a concern due to adversarial attacks. In order to tackle this issue, researchers have developed sustainable defense strategies, but certain challenges remain. These challenges involve transferability, higher computing costs, and adaptability. Striking a balance between accuracy and robustness is difficult, as defense mechanisms often come with trade-offs between the two. Real-world situations demonstrate the practical implications of sustainable adversarial AI. For example, it improves the security of self-driving vehicles, enhances the accuracy of medical imaging diagnoses, and incorporates AI-driven defenses into network intrusion detection and phishing detection systems. It is crucial to consider ethical aspects throughout this process. Future trends in adversarial AI research for cybersecurity will involve ensemble defense mechanisms, adversarial learning from limited data, and hybrid attacks. By embracing the evolving landscape, researchers and practitioners can develop sustainable AI systems that are more secure and resilient, effectively countering adversarial threats. 2023 IEEE. -
Machine Learning Approaches for Suicidal Ideation Detection on Social Media
Social media suicidal ideation has become a serious public health issue that requires creative solutions for early diagnosis and management. An extensive investigation of machine-learning techniques for the automated detection of suicidal thoughts in internet postings is presented in this research. We start off by talking about the concerning increase in information on social media about mental health issues and the pressing need to create efficient monitoring mechanisms. The research explores the several methods used to identify the subtleties of suicidal thought conveyed in text, photographs, and audio-visual information. These methods include sentiment analysis, natural language processing, and deep learning models. We look at the problems with unbalanced data, privacy issues, and the moral ramifications of keeping an eye on user-generated material. We also go over the research's practical ramifications, such as the creation of instruments for real-time monitoring and crisis response techniques. Through comprehensive experiments and benchmarking, we demonstrate the potential of machine learning in providing timely support for those in need, thereby reducing the impact of suicidal ideation on society. 2023 IEEE. -
Enhanced Automated Online Examination Portal Using Convolutional Neural Network
In recent years, the digital evolution of education has significantly shaped the landscape of learning, steering it away from traditional classroom settings towards more agile e-learning platforms. This shift has underscored the urgency for comprehensive online examination systems, tailored to meet the unique challenges and demands of virtual education. Online learning platforms have seen a rapid rise in popularity, given their flexibility, cost-effectiveness, and capability to cater to learners worldwide. Such a widespread audience brings along the challenge of conducting exams without the constraints of geography and scale. Traditional examinations, with their manual paper based formats, fail to fit within this digital mold due to their logistical challenges and inefficiencies. Consequently, an online examination system not only introduces convenience but also operational efficiency, eliminating many of the logistical nightmares associated with manual exams. While existing tools might provide online testing capabilities, the integration of Artificial Intelligent driven proctoring in this portal elevates the standards of academic integrity to unprecedented levels. The main aim of this article is to create online test platform with the support of Artificial Intelligence technology. The result detect the malpractice activity and electronic device usage detection while online examination. 2023 IEEE. -
QSAR Approach for Drug Discovery Targeting the Glucagon Receptor Using Machine Learning
Metabolic disorders like type 2 diabetes are increasing day by day so the study focusing drug discovery of glucagon receptor has become important.One of the method to study the binding strength between chemical compounds is Quantitative Structure-Activity Relationship (QSAR) which is discussed in this paper.We gathered a curated dataset of glucagon receptor ligands from the ChEMBL bio activity dataset and studied the physical and chemical properties of the molecules using factors like molecular weight and logarithm of the partition coefficient.Then Random forest regression model was applied for prediction of the binding strength of ligands. The efficiency information of ligand was extracted which contributed to study of the molecular features concerning the activity of glucagon receptor in a much easier manner. These findings highlight the potential of QSAR in elucidating the key determinants of ligated-receptor interactions and guiding the rational design of novel glucagon receptor modulators. The integration of computational approaches with experimental validation holds promise for accelerating the development of effective therapies for metabolic disorders, addressing unmet clinical needs in this field. 2023 IEEE. -
Design and Implementation of an Optimized Mask RCNN Model for Liver Tumour Prediction and Segmentation
Segmentation of liver tumour is a tedious job due to their large variation in location and closeness to nearby organs. In this research, a novel Mask RCNN prototype is developed which uses ResNet-50 model. The architecture utilizes the masked location of convolution neural network to precisely detect liver tumours by recognizing liver sites to deal with changes in liver and CT snaps with distinct metrics. The preprocessed CT scans are subjected to ResNet-50 model. The data samples used here comprises 130 instances recorded from several clinical sites that are publicly available on the LiTS weblink. The designed model upon deployment generates a promising outcome thereby obtaining a DSC of 0.97%. Thus, we can conclude that the developed model is capable enough to accurately assess liver tumours and thus help patients in early diagnosis. 2023 IEEE. -
Wood Type Identification via Neural Networks and Spectral Analysis: An Advanced Algorithmic Solution
Forestry management, woodworking, and manufacturing need wood type identification. This study introduces a neural network-spectral analysis technique for accurate and automatic wood type detection. Principal Component Analysis (PCA) is used to extract features from a heterogeneous collection of wood spectral signatures after training a neural network. The algorithm's 94.2% accuracy on a testing dataset shows its ability to distinguish wood kinds.The model's confusion matrix shows it can recognise closely related wood species with few misclassifications. The neural network's precision, recall, and F1 score prove its wood classification accuracy. With PCA highlighting classification characteristics, spectral analysis helps the algorithm succeed.The method is useful for forestry management and woodworking quality control. The non-destructive technology provides in-situ wood type detection, addressing environmental and conservation issues. The study explores ramifications, constraints, and future algorithm modification and application in real-world contexts.Neural networks and spectral analysis provide a strong, efficient, and non-destructive wood type detection solution. The hopeful results represent a major advance in wood science and current computer methods, with applicability across sectors. 2023 IEEE. -
An Abstractive Text Summarization Using Decoder Attention with Pointer Network
Nowadays, large amounts of unstructured data are currently trending on social media and the Web. Text summarising is the process of extracting pertinent information in a concise manner without altering the content's core meaning. Summarising text by hand requires a lot of time, money, and effort. Although deep learning algorithms are commonly applied in abstractive text summarization, further research is clearly needed to fully understand their conjunction with semantic-based or structure-based approaches. The resume dataset is taken for this research work, which is gathered from Kaggle and the dataset includes 1,735 Resumes. This paper presents a unique framework based on the combination of semantic data transformations and deep learning approaches for improving abstractive text summarization. In an attempt to tackle the problem of unregistered words, a solution called Decoder Attention with Pointer Network (DA-PN) has been introduced. This method incorporates the use of a coverage mechanism to prevent word repetition in the generated text summaries. DA-PN is utilized for protecting the spread of increasing errors in generated text summaries. The performance of the proposed method is estimated using the evaluation indicator Recall Oriented Understudy for Gisting Evaluation (ROUGE) and attains an average of 26.28 which is comparatively higher than existing methods. 2023 IEEE. -
A Self-Attention Bidirectional Long Short-Term Memory for Cold Start Movie Recommendation Models
Movie recommendation systems are useful tools that help users find relevant results and prevent information overload. On the other hand, the user cold-start issue has arisen because the system lacks sufficient user data. Furthermore, they are not very scalable for use in extensive real-world applications. One of the key strategies to address the sparsity and cold-start problems is to leverage other sources of information, including item or user profiles or user reviews. Processing client feedback is typically a challenging process that involves challenging the interpretation and analysis of the textual data. Thus, this research implements an efficient deep learning-based recommendation architecture. Following the acquisition of textual data from the Amazon product reviews database, stop word removal, lemmatization, and stemming techniques are applied to the data pre-processing which eliminate inconsistent and redundant data, facilitating the process of interpreting and utilising data. Then, the Term Frequency-Inverse Document Frequency (TF-IDF) method is applied to extract the feature values from the pre-processed text data. The extracted feature values are fed to the Self-Attention Bidirectional Long Short-Term Memory (SA-BiLSTM) that utilises the matrix factorization method framework's information sources. The SA-BiLSTM model obtained 95.93% of recall, 94.76% of precision, and 97.84% of accuracy on the amazon product reviews database. 2023 IEEE. -
Dynamic Behaviour Analysis of Multi-Cell Battery Packs: A Simulation Study
In the era of IoT understanding the dynamic behavior of a Lithium-ion Battery Management System (BMS) has become gradually more important. This research investigates the dynamic behaviour of a six-cell Lithium-ion Battery Management System (BMS) through simulation. The study employs a comprehensive model encompassing key battery parameters, including cell capacity, voltage limits, temperature thresholds, and charge/discharge characteristics. Additionally, state variables such as State of Charge (SOC), State of Health, and State of Function are integrated to capture the battery's internal dynamics. The simulation incorporates a sinusoidal current profile to emulate realistic operating conditions. Notably, Coulomb counting is employed for SOC estimation, and protective measures against overvoltage, undervoltage, and overcurrent are implemented. The study also addresses balancing strategies and communication interfaces within the BMS. The results reveal nuanced interactions between voltage, temperature, SOC, and current, offering insights into the intricate behaviour of the battery system under dynamic conditions. This research not only advances our understanding of BMS functionality but also lays a crucial foundation for the evolution of battery technology and energy management systems in the IoT landscape. The Institution of Engineering & Technology 2023. -
Enhanced Social Media Profile Authenticity Detection Using Machine Learning Models and Artificial Neural Networks
Fake engagement is one of the main issues with online networks or ONSs, which are used to artificially boost an account's popularity, this study examines the effectiveness of seven sophisticated Machine Learning Algorithms, Random Forest Classifier, Decision Tree Classifier, XGBoost, LightGBM, Extra Trees Classifier, and SVM, and got 93% accuracy in Decision Tree Classifier. In order to solve overfitting issues and improve model resilience, the paper proposes Generative Adversarial Networks (GANs) and uses K-Fold Cross-Validation. Furthermore, design a Gan-ANN model that combines Batch Normalization and Artificial Neural Networks (ANN) with GAN-generated synthetic data is investigated. The enhanced dataset seeks to strengthen model performance and generalization when combined with cutting-edge modeling methods. This study aims to improve model scalability, predictive accuracy, and dependability across different machine learning paradigms. 2023 IEEE. -
Early Detection of Cervical Cancer using Machine Learning Classifiers for Improved Diagnosis in Underserved Regions
One of the incurable diseases that affect women is cervical cancer. It is brought on by a protracted infection of the skin and the vaginal mucous membrane cells. The Human Papilloma Virus (HPV), is the main factor causing aberrant cell proliferation in the area around the cervix. There are no symptoms present when the illness first appears. Early detection of this malignancy may be used to prevent death. People in less developed countries cannot afford to periodically examine themselves due to a lack of awareness, poor medical infrastructure, and expensive medication. The EDA technique is applied to examine the data and understand its characteristics. Machine Learning algorithm has been used to diagnose cervical cancer. In order to spot the existence of cervical cancer, five machine learning classifiers are utilized, the algorithms to begin earlier. The Logistic Regression classifier's results validate the correct stage prediction. 2023 IEEE. -
EEG Signals Acquisition and Processing of Mental Tasks for Controlling Smart Systems
With the help of this biosignal, we can create various kinds of interfaces. These interfaces were especially used for paralyzed individuals who have problems in the normal bio-sensor channels. Electroencephalogram (EEG) signals play a vital role in semi-paralyzed and fully paralyzed individuals to make communication with their caretakers. EEG signals are used not only for paralyzed individuals, nowadays most of home appliances are designed due to their reliability and accuracy. EEG based Smart devices are very easy to operate because signals are generated by the human brain automatically whenever the task is performed by humans. Through this research, we discussed the basic methodology needed for interfacing the EEG signals with smart devices. From this paper, most of the researchers will be going to know how to connect the EEG signals with the smart systems to control the external devices and induce the researchers to apply the EEG signals in real time. 2023 IEEE. -
A Comprehensive Study of Blockchain Technology Based Decentralised Ledger Implementations
Information management and decentralized, secure transactions are made possible by the groundbreaking idea of blockchain technology. Blockchain has attracted considerable attention from a wide range of businesses as a result of the rising popularity of cryptocurrencies like Bitcoin and Ethereum. The primary goal of this paper is to present a thorough examination of decentralized ledger systems based on blockchain technology. It examines at the basic ideas, underpinnings, and real-world uses of blockchain in many industries. The ability to scale, effectiveness, privacy, security, seamless integration, governing scenarios, and consumption of energy are just a few of the technical factors that are looked at in relation to the deployment of blockchain technology. In this research paper we also pinpoint forthcoming developments and future prospects for the blockchain industry like Solutions for scalability at the subsequent layer, protocols for seamless integration, integration with the Internet of Things (IoT), apps for decentralized finance (DeFi), and digital currencies that are issued by central banks (CBDCs) are a few examples. 2023 IEEE. -
Automation using Artificial Intelligence in Business Landscape
The integration of Artificial Intelligence (AI) with automation has sparked a remarkable transformation in the contemporary business landscape, promising elevated efficiency and quality. However, this convergence encounters multifaceted challenges, notably in the adoption of recent AI techniques such as deep learning, reinforcement learning, and natural language processing. These techniques, while potent, grapple with challenges in data quality, interpretability, and ethical considerations. In this study, we aim to delineate the intricate interplay between AI and automation, illuminating their collective potential to augment operational efficiency and confer a competitive advantage. Through a comprehensive review, we will explore the effective integration of these technologies, navigating hurdles such as data bias, system compatibility, and human-machine collaboration. Here, the primary research objective is to provide insights on optimizing the outcomes by synergizing AI and automation while addressing the inherent challenges, ultimately fostering sustainable and impactful implementations in organizational frameworks. 2023 IEEE. -
Explaining Autism Diagnosis Model Through Local Interpretability Techniques - A Post-hoc Approach
In this era of machine learning and deep learning algorithms dominating the Artificial Intelligence (AI) world, the trustworthiness of these black box models is still questionable. Life-caring sectors like healthcare and banking make use of these black box models as assistance in critical decision-making processes, but the degree of reliability of these decisions is still uncertain. This is because these black box models will not reveal the causation of the predicted outcome. However, creating an interpretable model that can explain the internal workings of these black box models can provide some reliable insights and trustable justifications for the predicted outcome. This study aimed to create an interpretable model for autism diagnosis which can give some trustable explanations for its predicted outcome. Using local interpretability methods such as LIME, SHAP, and Anchors the predicted outcome for each instance is explained well with some standard visual representations. As a result, this study developed an interpretable autism diagnosis model with an accuracy rate of 91.37% and with good local model explanations. 2023 IEEE. -
Time Series Forecasting of Stock Market Volatility Using LSTM Networks
Forecasting stock market volatility is a pivotal concern for investors and financial institutions alike. This research paper employs Long Short-Term Memory (LSTM) networks, a potent class of recurrent neural networks, to predict stock market volatility. LSTM networks have proven adept at capturing intricate temporal dependencies, rendering them a fitting choice for time series data analysis. We commence by elucidating the notion of stock market volatility and its profound significance in financial decision-making. Traditional methodologies, such as GARCH models, exhibit shortcomings in deciphering the convoluted dynamics inherent in financial time series data. LSTM networks, with their capacity to model extended temporal relationships, present an encouraging alternative. In this study, we assemble historical stock price and trading volume data for a diverse array of assets, diligently preprocessing it to ensure its aptness for LSTM modeling. We systematically explore various network architectures, hyperparameter configurations, and input features to optimize the efficacy of our models. Our empirical investigations decisively underscore the supremacy of LSTM networks in capturing the subtleties of stock market volatility compared to conventional techniques. As the study progresses, we delve deeper into the complexities of LSTM network training, leveraging advanced techniques such as batch normalization and dropout to fortify model resilience. Moreover, we delve into the interpretability of LSTM models within the context of stock market forecasting. 2023 IEEE. -
KESMR: A Knowledge Enrichment Semantic Model For Recommending Microblogs
In today's world, there's an enormous amount of information available on the Internet. Because of this, it's become really important to come up with better and smarter ways to search for things online. The old-fashioned methods, like just matching certain words or using statistics, don't work so well anymore. They often suggest web pages that are irrelevant. As the Semantic Web keeps getting bigger, it needs algorithms for the best fit. In this paper, a way to measure how different the words used for web search. This helps in suggesting the most relevant web pages. A special algorithm that can change its settings. Our proposed method demonstrates 94% accuracy. 2023 IEEE. -
Twitter Sentiment Analysis and Emotion Detection Using NLTK and TextBlob
On an average, approximately 7000 tweets are communicated each second and in total it piles up to around 300 billion tweets every year. Society are free to contribute their opinions on public platform and hence it acts as a reliable interface to assess society ongoing viewpoint and attitude over any matter or event. Consumers very often make use of social media to exchange their views about anything. Business may get domain for enhancement and smooth interpretation of the behavior of people regarding various facts through opinion mining. Thus to carry out this mining of opinions on social media interface, textual categorization with language analysis is of great help. With the help of NLP token tool, phrases can be divided into various word series after dropping stop phrases. Larger tweets tokenizing and classifying into distinct labels is a concern. Thus, the main objective of this framework is to process the tweets based on specific keywords given by user, categorize these phrases into negative, positive and neutral ones. TextBlob module assists users and developers to interpret user sentiments about a news. This research tries to give suggestion a textual opinion assessment on social media samples utilizing the NLTK and TextBlob modules. 2023 IEEE. -
Comparative Analysis of Maize Leaf Disease Detection using Convolutional Neural Networks
Worldwide, maize is a significant cereal crop for crop productivity, identifying diseases in the plant's leaves is essential to raise a good crop. Deep learning methods that have been used in recent years to precisely identify and categorize these serious diseases, offering a non-destructive and effective way to find maize leaf ailments. In order to detect maize leaf disease, this paper suggests using three well-liked deep learning models: VGG16, Inception V3, and EfficientNet. The models were trained and assessed using a datasets of 4000 images of three distinct maize leaf diseases and a healthy class. All three models had high accuracy rates, according to the results, though EfficientNet outperformed the other two models. The suggested method can detect and track diseases in maize crops with high accuracy and can be applied practically. It can accurately classify various diseases. The study also demonstrates that deep learning models can offer a trustworthy and effective solution for detecting crop diseases, which can aid in lowering crop losses, raising crop yields, and enhancing food security. 2023 IEEE. -
Detection and Robust Classification of Lung Cancer Disease Using Hybrid Deep Learning Approach
Effective lung cancer diagnosis and treatment hinge on the early detection of lung nodules. Various techniques, such as thresholding, pattern recognition, computer-aided diagnostics, and backpropagation calculations, have been explored by scientists. Convolutional neural networks (CNNs) have emerged as powerful tools in recent times, revolutionizing many aspects of this field. However, traditional computer-aided detection systems face challenges when categorizing lung nodule detection. Excessive reliance on classifiers at every stage of the process results in diminished recognition rates and an increased occurrence of false positives. To address these issues, we present a novel approach based on deep hybrid learning for classifying lung lesions. In this study, we explore multiple memory-efficient and hybrid deep neural network (DNN) architectures for image processing. Our proposed hybrid DNN significantly outperforms the current state-of-the-art, achieving an impressive accuracy of 95.21%, all while maintaining a balanced trade-off between specificity and sensitivity. The primary focus of this research is to differentiate between CT scans of patients who have early-stage lung cancer and those who do not. This is achieved by utilizing binary classification networks, including standard CNN, SqueezeNet, and MobileNet. 2023 IEEE.