IFIP Open Access Digital Library
Not a member yet
22614 research outputs found
Sort by
Augmenting the One-Worker-Multiple-Machines System: A Softbot Approach to Support the Operator 5.0
No full textPart 3: Inclusive Work Systems Design: Applying Technology to Accommodate Individual Worker’s NeedsInternational audienceIndustry 4.0/5.0 workplaces are characterized by humans surrounded by massive digitalization, huge data generation, and data-driven management. However, this brings more complexity to the operators as they are exposed to vast amounts of data to reason about as well as to many situations of overwhelming cognitive load, leading them to potentially less assertive and stressful decision-making. This becomes challenging when operators should manage two or more machines simultaneously as in a ‘One-Worker-Multiple-Machines’ (OWMM) working environment, including critical processes and equipment. This paper proposes a softbot approach to address these issues devising an OWMM smart cockpit environment where an intelligent softbot supports an operator in several production situations. A software prototype was developed to show the potential and benefits of the softbot approach in OWMM environments
Waste Segregation Using Deep Learning Model
No full textPart 4: SDG 11 Sustainable Cities and CommunitiesInternational audienceWaste segregation has become one of the crucial tasks which can in turn contribute to efficient recycling of wastes. Manual waste segregation techniques have a lot of issues and also time and efficiency constraints. All these have led to the need for a smart waste segregation system. The main objective is to develop a waste segregation system using a deep learning model built on YOLOv8 architecture, deployed on raspberry pi. The system focuses to classify the wastes to glass, metal, plastic and others by training the model with suitable dataset for each class. The experimental results show that classification model provides high accuracy and fast prediction making it a suitable tool for waste segregation
Time Series Forecasting for COVID-19 Confirmed Cases Using Transformer Based Stacked LSTM Model
No full textPart 1: SDG 3 Good Health and Well-BeingInternational audienceThe worldwide impact of COVID-19 necessitates accurate forecasting for informed decision-making by governments and health entities. This study proposes neural network architectures for time series forecasting of confirmed COVID-19 cases based on historical data. The models, including variants of LSTM networks, CNN-GRU hybrids, RBMs, DBNs, and self-attention mechanisms, are trained on comprehensive datasets of confirmed cases, deaths, recoveries, and other variables across various countries. Comparative analysis indicates that CNN-GRU and LSTM models exhibit superior performance in forecasting accuracy, surpassing RBM, DBN, and self-attention models across metrics like Mean Absolute Error and Root Mean Squared Error. This research emphasizes the viability of employing advanced neural networks for precise COVID-19 spread predictions, pivotal in shaping public health policies during global crises
“Probably Put Some Sort of Fear in”: Investigating the Role of Heuristics in Cyber Awareness Messaging for Small to Medium Sized Enterprises
No full textPart 1: Awareness and EducationInternational audienceCyber-attacks are increasing at an exponential rate, targeting organisation irrespective of size. Small to medium sized enterprises (SMEs) are particularly vulnerable yet often lack cybersecurity awareness. This entails that an individual or organisation becomes aware of the cyber threats they face in addition to the protective actions and behaviours they can take. Despite the positive intentions of current cybersecurity awareness initiatives, there is a lack of adoption by SMEs. To better understand the situation this study explores SME owner or manager perceptions of cybersecurity awareness messages, leveraging psychological heuristics and message framing. Empirical data was collected through interviews with 16 participants representing SMEs in the North-East of England. Findings reflect that the framing of messages towards fear is more accepted by SMEs as opposed to positivity messages. Moreover, heuristics of self-efficacy and cost are seen to instil a desire to comply with cyber security behaviours. However, not all SMEs could agree on an approach thus suggesting that SMEs require bespoke messaging relating to the businesses and the owner
Sentiment Analysis of Various Ride Sharing Applications Reviews: A Comparative Analysis Between Deep Learning and Machine Learning Algorithms
No full textPart 3: Applications of MLInternational audienceThe exponential growth of ride sharing platforms in developing countries has significantly transformed the landscape of urban transportation. It is crucial for service providers to comprehend customer feedback and opinions in order to enhance the user experience and for passengers to select a safe and secure ride. A comprehensive sentiment analysis was conducted on user evaluations of ride-sharing platforms such as Grab, Uber, Indrive, Pathao, Jatri, and Obhai. An aggregate of 12052 data points (Negative, Positive, Neutral) were extracted from ride sharing reviews available on the Google Play Store. By utilizing cutting-edge deep learning models (LSTM, GRU, BiLSTM, and BiGRU) in addition to traditional machine learning models (Random Forest, Decision Tree, Gradient Boosting Classifier, Logistic Regression, KNN, Naive Bayes, SVM, AdaBoost, and LightGBM), our research endeavors to offer substantial insights into the determinants of customer satisfaction. Support Vector Machine is the most effective machine learning algorithm in this context, attaining an accuracy rate of 76.70%. Bidirectional Gated Recurrent Unit emerged as the best-performing model in the domain of deep learning, attaining an exceptional accuracy rate of 94.97%, this algorithm demonstrates the efficacy of deep learning methodologies in the context of sentiment analysis. Our research indicates that deep learning algorithms demonstrate significant superiority when compared to machine learning algorithms
Heart Attack Prediction Using Big Data Analytics
No full textPart 2: Data AnalyticsInternational audienceAcute myocardial infarction, sometimes known as a heart attack, is among the most fatal conditions that patients encounter. Large-scale data analysis, comparison, and mining for insights that can be applied to forecast, prevent, manage, and cure chronic illnesses like heart attacks are crucial to effectively treating cardiovascular illness. With tremendous efficacy, big data analytics—which is widely recognized in the corporate world for its application in regulating, comparing, and managing massive datasets—can be utilized to forecast, prevent, manage, and treat cardiovascular disease. Large-scale datasets can be analyzed using big data technologies like Hadoop, data mining, and visualization
MLEE: Event Extraction as Multi-label Classification Task at Token Level
No full textPart 1: Machine LearningInternational audienceEvent Extraction is an important task in natural language understanding, which aims to identify event trigger of pre-defined event types and their arguments of specific roles, has attracted a lot of attention from industry and academia. The previous works failed to address some issues, including error propagation problem, overlap and nest problem, and high complexity of model. This work proposes a novel model MLEE, which models Event Extraction task as Multi-Label classification task at token level, and processes the extraction task in a joint paradigm, can help solving issues mentioned above. The experiment verifies our model’s effectiveness. Empirical results on DuEE and FewFC shows that MLEE outperforms previous best model, pushing trigger extraction F1 to 85.03% (+4.45%), argument extraction F1 to 78.85% (+2.72%) on DuEE, pushing trigger extraction F1 to 76.69% (+1.63%), argument extraction F1 to 76.53% (+5.27%) on FewFC
Empowering Medical Image Analysis: Unveiling Anomalies Through GANs and BiGAN’s Models
No full textPart 2: Applications of AI/ML in Image ProcessingInternational audienceIn many researches, the field of medical image analysis has made noteworthy strides, thanks to the implementation of sophisticated machine learning methods. This project, titled “Enhancing Medical Image Analysis: Uncovering Abnormalities through Innovative Models”, stands as a modern game powerup to utilize rapidly progressing computer algorithms known as Generative Adversarial Networks (GANs) and Bidirectional GANs (BiGANs) for the automated identification and categorization of anomalies within medical images. Medical image analysis plays a crucial role in diagnosing and treating a range of health conditions, in this case particularly of 2 types, Bacterial and Viral. However, the precision and accuracy of image analysis often depend upon the expertise and execution of radiologists, which can introduce the potential for human errors and result in time-consuming processes. To tackle these challenges, this project introduces a fresh approach that makes use of GANs and BiGANs to improve the identification and categorization of abnormalities in medical images, thereby providing valuable insights to healthcare professionals and ultimately enhancing patient diagnosis and treatment outcomes
Achieving Adaptive Safety via Trust Building in Autonomous Ecosystems
No full textPart 4: Cybersecurity and SafetyInternational audienceThe evolving landscape of Autonomous Cyber-Physical Systems is progressing towards cooperative ecosystems. These complex structures, characterized by dynamic interactions among various Autonomous Systems, offer heightened autonomy and adaptability but pose substantial challenges in ensuring Safety. The swift development of autonomous driving underlines the urgency to address these issues. Existing Safety assurance methods, while effective on an individual level, struggle to encompass the complexities of dynamic ecosystems. To bridge this gap, this paper advocates for adaptive Safety mechanisms informed by Trust and trustworthiness between ecosystem members and proposes a path towards a method ensuring adaptive Safety
Predicting Song Popularity Through Machine Learning and Sentiment Analysis on Social Networks
No full textPart 2: The 13th Workshop on “Mining Humanistic Data” (MHDW)International audienceIn this paper, we delve into the intricate relationship between technology, music, and success. Our research focuses on leveraging the capabilities of machine learning algorithms to forecast the success and popularity of songs. By combining audio features, social media data, and emotion analysis, our machine learning-based model demonstrates its capability as a reliable tool for artists, amateurs, and music industry stakeholders to make informed and targeted decisions. Our analysis encompassed not only the inherent audio characteristics of each track but also the extent of audience engagement and emotional response of the fans. Through the integration of textual emotion analysis, we quantified the reactions of the audience and performed an emotion labelling, leveraging our social information to uncover the emotions surrounding each track’s release. The experimental results show that the feature data from which the classification algorithms were trained, unfolded to be qualitative and precise, while they also show that the use of the social media features improves the classifiers’ performance in predicting the popularity of the tracks