Browse Items (11808 total)
Sort by:
-
An Enhanced Deep Learning Model for Duplicate Question Detection on Quora Question pairs using Siamese LSTM
The question answering platform Quora has millions of users which increases the probability of questions asked with similar intent. One question may be structured in two different ways by two users, and answering similar questions repeatedly impacts user experience. Manual filtration of such questions is a tedious task, so Quora attempts to detect and remove these duplicate questions by using the Random Forest Model, which is not completely effective. As Quora contains question answers in the form of text data, different Natural Language Processing techniques are used to transform the text data into numerical vectors. In this research, the log loss metric acts as the primary metric to evaluate different models. The primary contribution is that the Siamese network is used to process two questions parallelly and find vectors representation of each question. The vectors computed by this method enables similarity detection which is more effective than existing models. GloVe word embedding is used to understand the semantic similarity between two questions. The random classifier is built as the base model and logistic regression, linear SVM and XGBoost model are used to reduce the log loss. Finally, a Siamese LSTM is proposed which reduces the loss dramatically. 2022 IEEE. -
An Efficient Deep Learning-Based Hybrid Architecture for Hate Speech Detection in Social Media
Social media has become an integral part of life as users are spending a significant amount of time networking online. Two primary reasons for its increasing popularity are ease of access and freedom of speech. People can express themselves without worrying about consequences. Due to lack of restriction, however, cases of cyberbullying and hate speeches are increasing on social media. Twitter and Facebook receive over a million posts daily, and manual filtration of this enormous number is a tedious task. This paper proposes a deep learning-based hybrid architecture (CNN + LSTM) to identify hate speeches by using Stanfords GloVe, which is a pre-trained word embedding. The model has been tested under different parameters and compared with several state-of-the-art models. The proposed framework has outperformed existing models and has also achieved the best accuracy. 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. -
Smart Home Activity Recognition for Ambient Assisted Living (AAL)
With the increasing age of an individual, the chances of being prone to chronic diseases like diabetes or non-curable diseases like Alzheimer's Syndrome or Parkinson's Syndrome increases. Due to the health issues, elderly must be accompanied by caretakers to monitor their well-being at all times. With growing responsibilities and work pressure, the family members may find it challenging to find a trustworthy caretaker. In such scenarios, an assisted living environment acts as a boon. A normal home embedded with different sensors to monitor an individual's well-being is called as Ambient Assisted Living(AAL). This living environment detects anomalous behaviour and recognizes human activities. In this research paper, a smart home activity recognition model is proposed and implemented using four machine learning algorithms using six different publicly available datasets. It has been observed that Random Forest machine learning algorithm shows the best accuracy on most of the dataset. The Electrochemical Society -
Masked Face Recognition and Liveness Detection Using Deep Learning Technique
Face recognition has been the most successful image processing application in recent times. Most work involving image analysis uses face recognition to automate attendance management systems. Face recognition is an identification process to verify and authenticate the person using their facial features. In this study, an intelligent attendance management system is built to automate the process of attendance. Here, while entering, a persons image will get captured. The model will detect the face; then the liveness model will verify whether there is any spoofing attack, then the masked detection model will check whether the person has worn the mask or not. In the end, face recognition will extract the facial features. If the persons features match the database, their attendance will be marked. In the face of the COVID-19 pandemic, wearing a face mask is mandatory for safety measures. The current face recognition system is not able to extract the features properly. The Multi-task Cascaded Convolutional Networks (MTCNN) model detects the face in the proposed method. Then a classification model based on the architecture of MobileNet V2 is used for liveness and mask detection. Then the FaceNet model is used for extracting the facial features. In this study, two different models for the recognition have been built, one for people with masks another one for people without masks. 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. -
MIST-based Tuning of Cyber-Physical Systems Towards Holistic Healthcare Informatics
The entire world seems shaken and disrupted since the strike of Covid-19 ever since its outbreak towards the end of 2019 and its continued perils. During this unprecedented event of the century, people's health emerged as the most vulnerable and affected area either directly or indirectly by the coronavirus and its new variants. Disrupting almost all spheres of life, patients' health and care systems required timely support from healthcare professionals to provide the needed medical advice on one hand and a prescriptive mechanism to avoid another impending casualty. Similarly, a proactive approach became desirable from the health ministry, pharmaceutical firms, medical insurance companies, and other stakeholders in fine-tuning their offerings to the patients as per the recommender systems. The devices to measure the vitals of a person, became more efficient and ergonomically sound with the advent of wearable gadgets. These devices monitored the physical activities of the user and transferred the vital signals wirelessly to any base computing device and cloud-based repositories. This mechanism, however, was reported to fail in addressing the issues with non-communicating or stand-alone devices that were used by the masses in developing countries including India. If the real-time data could be used from these devices, the healthcare diagnosis and analysis of a patient's medical condition could have taken a progressive dimension with the addition of missing data points. This research thus aims to fill the information gap and proposes a transforming approach towards existing non-communicating devices used to measure the vitals like blood pressure, oxygen level, blood sugar, etc. The proposed MIST-based Cyber-Physical System shall create extensive scalability towards the retrieval of the vital details from the devices which were otherwise captured offline previously and were unused at multiple critical points of healthcare processes. 2022 IEEE. -
Research Initiative on Sustainable Education System: Model of Balancing Green Computing and ICT in Quality Education
Green Computing Practices (GCP) convey the revolutionary changes of the modern education system. The education system is transforming into a hybrid mode of operations in effective teaching and learning procedure. In the modern era, computer devices are playing a foremost role in performing ICT based teaching and learning (ICT-BTL). The GCP and ICT-BTL are the creative and innovative practices that can ensure the eco-friendly enactment and safeguard from various harmful environmental impacts. The motive of projecting the present research outcome is to address the impact of GCP on ICT-BTL activities. The creative and innovative practices of ICT-BTL support the implementation of GCP towards a sustainable education system. A sustainable education system interconnects the teachers, learners, institutions, and industrial experts through eco-friendly electronic and computer devices that ensure maximum efficiency in education with minimum environmental impacts. 2022 IEEE. -
Removal of arsenic using ecofriendly egg shell and black toner powder
This work is primarily focused on the study of the possible usage of ecofriendly black toner powder and egg shell powder as adsorbent material for the removal of arsenic from industrial effluent. Batch experiments were conducted by varying the concentration, size of the reinforcement particles, time and its pH value. The optimal pH for the effective removal of arsenic was found to be 7. The size of the particles played a significant role in removing the arsenic. Smaller size particles outperformed the bigger size particles and the joint action of intra-particle transfer and pore diffusion mechanism played a major role in the removal of arsenic. 2022 -
A Comparative Analysis of Competition Law Regimes with the Increase of E-Commerce in India and U.S.A
The growth in analytics and cloud technologies has provided an interface where it is more interactive and approachable for the consumer to decide about purchases and varieties. The authors in this paper will be addressing the existence of anti-competitive practices in India, US and provide a comparative study of the enforceability of Competition laws in these countries respectively. India is primarily considered as one of the lucrative markets with highest usage of mobile phones and data and growing demand for the same, the new entrants in the market are finding it difficult with the anti-competitive aspects for instance unfair practices by gate keepers. The authors will research on the need to promote economic growth post pandemic and the necessary steps to be incorporated in such promotions so as to increase the demand and supply but at the same time maintain the competition. The Electrochemical Society -
Classification of Vitiligo using CNN Autoencoder
Precise recognition of skin ailment is a time-consuming procedure even for Professionals. With the invention of deep learning and medical image processing, the identification of skin disease is possible in a time-efficient manner and accurately. Autoencoder is the generative algorithm but in the proposed work it is used as a generator and as well as a classifier. In this work, a Convolutional (CNN) autoencoder was used to classify the skin disease Vitiligo. In this work encoding and decoding layers were used but in the last layer in place of reproducing the original image, the classification layer was used to classify the image. The proposed work gave training accuracy of 87.71 % whereas validation accuracy was 90.16%. 2022 IEEE. -
Reconceptualizing Empowerment And Autonomy: Ethnographic Narratives From A Self Help Group In South India
The paper revisits academics' conceptualizations of women empowerment as stopping short of autonomy. It departs from the general observation that women empowerment movements by and large have failed to translate the new agency of women outside the domains of socio economy; that women empowerment movements' capacity to re-engage with patriarchal structures and ideologies is seriously contained. Through an ethnography of Kudumbashree, an SHG in the South Indian state of Keralam, we question the neat distinctions between empowerment and autonomy that prevail in the academic common sense. The transition of agency from the economic to the political domain is a subtle enterprise and is mediated by a number of factors including the economic independence, decision making capability and political participation. Socio -economic - political implications of women empowerment could be the first step in challenging and overcoming the relations of oppression in any society. The stereotypical assumptions can be negotiated by solely apportioning responsibilities and re-engaging with the system through everyday practices. The nuances of empowered women's re-engagement with local gender/power regimes lead to changes at the conceptual level that cuts beyond the individual and group level material transformations. The Electrochemical Society -
Automated Fetal Brain Localization, Segmentation, and Abnormalities Detection Through Random Sample Consensus
The ability to detect and identify prenatal brain abnormalities using magnetic resonance imaging (MRI) is critical, as one in every 1000 women is pregnant with one. The brain is abnormal. Detection of embryonic brain abnormalities at an early stage machine learning techniques can help you increase the quality of your data. Treatment planning and diagnosis according to the literature that the majority of the research done in order to classify brain abnormalities in the term "very early age" refers to preterm newborns and neonates, not fetal development. However, studies of prenatal brain MRI imaging have been published and compared these images to the MRI scans of newborns to identify a non-fetal aberrant behavior in neonates. In this case, a pipeline procedure, on the other hand, is time-consuming. In this research, a machine learning-based pipeline process for fetal brain categorization (FBC) is proposed. The classification of fetal brain anomalies at an early stage, before the baby is delivered, is the paper's key contribution. The proposed approach uses a flexible and simple method with cheap processing cost to detect and categorize a variety of abnormalities from MRI images with a wide range of fetal gestational age (GA). Segmentation, augmentation, feature extraction, and classification and detecting anomalies of the fbrain are different phases of the recent method. 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. -
Digital Forensics Investigation for Attacks on Artificial Intelligence
The new research approaches are needed to be adopted to deal with security threats in AI based systems. This research is aimed at investigating the Artificial Intelligence (AI) attacks that are malicious by design. It also deals with conceptualization of the problem and strategies for attacks on Artificial Intelligence (AI) using Digital Forensic tools. A specific class of problems in Adversarial attacks are tampering of Images for computational processing in applications of Digital Photography, Computer Vision, Pattern Recognition (Facial Mapping algorithms). State-of-the-art developments in forensics such as 1. Application of end-to-end Neural Network Training pipeline for image rendering and provenance analysis, 2. Deep-fake image analysis using frequency methods, wavelet analysis & tools like - Amped Authenticate, 3. Capsule networks for detecting forged images 4. Information transformation for Feature extraction via Image Forensic tools such as EXIF-SC, Splice Radar, Noiseprint 5. Application of generative adversarial Networks (GAN) based models as anti-Image Forensics [8], will be studied in great detail and a new research approach will be designed incorporating these advancements for utility of Digital Forensics. The Electrochemical Society -
Self Risk Assessment Model Embedded with Conversational User interface for Selection of Health Insurance Product
In this research, we propose a dynamic model that works through Human-Computer Interaction to facilitate a smooth customer experience for health insurance prospects. The model facilitates the prospects to self assess their health risks. The integration of Conversational User interface, such as Mobile User Interface, Graphic User Interface and Bots with transcoder permits seamless use of the model by any category of prospects, irrespective of their language. Moreover, the model also helps the visually impaired person to interact without any hassle with the presence of a transcoder that permits conversion of text into speech and vice versa. The learner model comprises of the Prospects' detail module and Risk Assessment modules. The Prospects' detail module collects data from the predefined list. The risk assessment module profiles and assesses the risk based on the data inputted in the Prospects' detail module. The risk assessment level module categorizes the level of risk as low, moderate or high for each prospect depending on the total risk exposure level. The total risk exposure level is computed based on the pre-defined threshold. This model aids the prospect in determining the risk level and thereby facilitates self-selection of health insurance policy, thus reducing over reliance on the insurer. This model helps the prospect to take an independent purchase decision. 2022 IEEE. -
Towards Sustainable Living through Sentiment Analysis during Covid19
Artificial intelligence is the process of the machine to perform with the simulation of human intelligence. Computing within the field of emotions paves the recognitions to sentiment analysis. Sentiment analysis is the method of capturing the emotions behind a text whether or not it's positive, negative or neutral. Sentiment Analysis (SA) or Opinion Mining (OA) is the process to provide computational treatment to unstructured data to categorize and identify the sentiments or emotions expressed in a piece of text. It combines Natural Language Processing Techniques and Machine Learning Techniques. This technology is additionally referred to as opinion mining or feeling computing. Sentiment Analysis uses the ideas of machine learning alongside an AI based process called NLP to extract and analyse the data, emotions, information from the text. This work explores the impact of social media during covid 19 and possible link between sustainable living and health care with the usage of sentiments. This paper address the sustainable development goal 3 (good health and wellbeing) of SDG 2030 and a possible way towards sustainable living through sentiment analysis. The Electrochemical Society -
Parametric effect of minimum quantity lubrication unit using RSM technique to improve the machinability of Inconel 718
In recent years, a rapid demand of superalloys has been seen in all industrial sectors. Few growing industries such as aerospace and biomedical industries are in need of this superalloy for fabrication of variety of products. Inconel 718 is one such superalloy which is being used for the manufacture of these productions due to high tensile strength, corrosion resistance, hardness and toughness. Due to these superior quality feature of this material friction is being seen at the tool-work interface region. This friction can be reduced by minimum quantity lubrication (MQL) unit which provides coolant at the right time. This paper discusses the minimum quantity lubrication unit used in computer numerical control (CNC) milling machine to improve the machinability of Inconel 718 by reduction of temperature at tool-work interface region and also the parametric effect using response surface methodology (RSM). Minimum quantity lubrication unit allows the cutting fluid to flow out of the nozzle at minimum speed to the cutting region which provides maximum volume of heat removal at very minimized usage of fluid. RSM technique is being implemented to improvise the experimental runs with proper way of extracting the readings and providing the observations. Instead of generating huge datas, RSM shows the accurate path of providing the data in a specified generative table. [copyright information to be updated in production process] 2022 -
Machine Learning Based Time Series Analysis for COVID-19 Cases in India
The World Health Organization declared the Coronavirus Infection, or COVID-19, to be widespread. One of the most appropriate methodologies for COVID-19 is time series analysis. The most appropriate technique for COVID-19 is time series analysis. It can be applied to Recognizing Information Patterns and Predicting Insights. The paper summarises the components of time series using the COVID-19 dataset for India as an example of one of the most important methodologies in predictive analytics. Time series models are chosen because they can predict future outcomes, comprehend prior outcomes, provide strategy recommendations, and much more. These common goalrists of temporal arrangement modelling do not differ significantly from those of cross-sectional or board data modelling. Machine Learning may be a well-known fact that it is an excellent technique for imagining, discourse, and standard dialect management for a large clarified accessible dataset. The results for confirmed, recovered, and death cases are presented in this study. 2022 IEEE. -
A Systematic Review of Challenges, Tools, and Myths of Big Data Ingestion
Each sector of the digital world generates enormous data as human life continues to transform. Areas like data analytics, data science, knowledge discovery in databases (KDD), machine learning, and artificial intelligence depend on highly distributed data which requires appropriate storage in a data lake. Collecting the data from different heterogeneous sources and creating a single lake of data is called data ingestion. Ironically, data ingestion has been treated as a less important stage in data analysis because it is considered a minor first step. There are several misconceptions in the data and analytics domain about data ingestion. The survey employed in this research presents a list of significant challenges faced by information technology (IT) industries during data ingestion. The available frameworks are compared in terms of standard parameters that are set against the existing challenges and myths. The findings from the comparison are compiled in a tabular format for easy reference. The paper places emphasis on the significance of data ingestion and attempts to present it as a major activity on the big data platform. 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. -
Predicting Stock Market Price Movement Using Machine Learning Technique: Evidence from India
The stock market is uncertain, volatile, and multidimensional. Stock prices have been difficult to predict since they are influenced by a variety of factors. In order to make critical investment and financial decisions, investors and analysts are interested in predicting stock prices. Predicting a stock's price entails developing price pathways that a stock might take in the future. ANN and mathematical Geometric Brownian movement technique were employed in this study to forecast a stock market closing price of Indian companies. The comparative analysis indicates that the Geometric Brownian Method is better than ANN in giving better MAPE and RMSE Values. 2022 IEEE. -
An AI-Based Forensic Model for Online Social Networks
With the growth of social media usage, social media crimes are also creeping sprightly. Investigation of such crimes involves the thorough examination of data like user, activity, network, and content. Although investigating social media looks quite straight forward process, it is always challenging for the investigators due to the complex process involved in it. Due to the immense growth of social media content, manual processing of data for investigation is not possible. Most of the works from this area provide an automatic model or semi-automated, and much of the contributions lacks the logical reasoning and explainability of the evidence extracted. Searching techniques like entity-based search and explainable AI add value to the quick retrieval within appropriate scope and explain the results to the court of law. This paper provides a model by adding these new techniques to the basic forensic process. 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. -
Improving Indoor occupancy estimation using a hybrid CNN-LSTM approach
Indoor Air Quality (IAQ) monitoring has been a significant research domain in energy conservation. Many energy resources are required to maintain the IAQ using airconditioning or other ventilation systems. Currently, the research works highly optimize an on-demand driven energy usage depending on the occupant present inside the building. In the last decade, numerous research works have evolved for such an optimization by installing sensors and predicting occupants using machine learning techniques. This research fails to deploy non-intrusive sensors and appropriate machine learning algorithms to predict the occupancy count. Advancement in neural network techniques termed deep learning has made significant performance in recognition and cognitive tasks. Thus, this paper proposes a hybrid deep learning model that stacks the convolutional neural network (CNN) and long short term memory (LSTM) to improve the prediction rate of the occupancy count. Experimentation has been carried out in real-time multivariate sensor data for the occupancy estimation and evaluated the performance in terms of accuracy, RMSE, MAPE, and coefficients of determinants. 2022 IEEE.