As the prevalence of deepfake videos continues to escalate, there is an urgent need for robust and efficient detection methods to mitigate the potential consequences of misinformation and manipulation. This abstract explores the application of Long Short-Term Memory (LSTM) networks in the realm of deepfake video detection. LSTM, a type of recurrent neural network (RNN), has proven to be adept at capturing temporal dependencies in sequential data, making it a promising candidate for analysing the dynamic nature of videos. (...) The research delves into the intricacies of utilizing LSTM architectures for the detection of deepfake videos, emphasizing the significance of understanding temporal patterns inherent in manipulated content. The proposed methodology involves preprocessing of video data, including the creation of high-quality training datasets and the application of data augmentation techniques to enhance model generalization. The training process and optimization strategies specific to LSTM networks are explored to achieve optimal performance in deepfake detection. Evaluation metrics such as accuracy, precision, recall, and F1 score are employed to assess the model’s effectiveness in distinguishing between genuine and manipulated content. The abstract also addresses challenges and limitations inherent in deepfake detection, including mitigating false positives and negatives, and discusses potential avenues for future research to enhance the robustness of LSTM-based detection systems. The findings of this research have implications for real-world applications, particularly in the context of social media platforms and video hosting services, where the integration of LSTM-based deepfake detection can contribute to a safer and more secure online environment. k. (shrink)

The results show that the hybrid CNN-LSTM model outperforms the individual models in terms of prediction accuracy and robustness, suggesting that combining convolutional layers with recurrent units is beneficial for capturing both spatial and temporal patterns in air quality data. This study demonstrates the potential of deep learning methods to offer real-time, accurate air quality forecasting systems, which can aid policymakers and urban planners in managing air pollution more effectively.

In This project focuses on employing machine learning techniques to forecast future events by retaining data over time intervals. Right now, the main tools for this job are things like Recurrent Neural Networks (RNN), including different types like LSTM & GRU. Employing machine learning methods has demonstrated effectiveness in analyzing stock prices, one can enhance decision-making accuracy. The adoption of ML techniques for stock value forecasting has surged in recent times. This research aims to construct an artificially intelligent model (...) specialized in estimating stock prices of specific companies. The focal point of this investigation revolves around employing LSTM a type of RNN, as the primary methodology . (shrink)

The Korea National Health and Nutrition Examination Survey (2020) reported that the prevalence of obesity (≥19 years old) was 31.4% in 2011, but it increased to 33.8% in 2019 and 38.3% in 2020, which confirmed that it increased rapidly after the outbreak of COVID-19. Obesity increases not only the risk of infection with COVID-19 but also severity and fatality rate after being infected with COVID-19 compared to people with normal weight or underweight. Therefore, identifying the difference in potential factors for (...) obesity before and after the pandemic is an important issue in health science. This study identified the keywords and topics that were formed before and after the COVID-19 pandemic in the South Korean society and how they had been changing by conducting a web crawling of South Korea's news big data using “obesity” as a keyword. (shrink)

: COVID-19 has proliferated over the earth, exposing mankind at risk. The assets of the world's most powerful economies are at stake due to the disease's high infectivity and contagiousness. The capacity of machine learning algorithms can estimate the amount of future COVID-19 cases, which is now considered a possible threat to civilization. Five conventional measuring models, notably LR, LASSO, SVM, ES, and LSTM, were utilised in this work to examine COVID-19's undermining variables. Each model contains three sorts of (...) expectations: the number of newly contaminated cases, the number of passings, and the number of recoveries. However, it is hard to anticipate the patients' real outcomes. To address the issue, a suggested approach based on long transient memory (LSTM) forecasts the number of COVID-19 cases in the next 10 days as well as the impact of preventative measures such as social isolation and lockdown on COVID-19 spread. (shrink)

This work demonstrates the implementation and use of an encoder-decoder model to perform a many-to-many mapping of video data to text captions. The many-to-many mapping occurs via an input temporal sequence of video frames to an output sequence of words to form a caption sentence. Data preprocessing, model construction, and model training are discussed. Caption correctness is evaluated using 2-gram BLEU scores across the different splits of the dataset. Specific examples of output captions were shown to demonstrate model generality over (...) the video temporal dimension. Predicted captions were shown to generalize over video action, even in instances where the video scene changed dramatically. Model architecture changes are discussed to improve sentence grammar and correctness. (shrink)

This work demonstrates the implementation and use of an encoder-decoder model to perform a many-to-many mapping of video data to text captions. The many-to-many mapping occurs via an input temporal sequence of video frames to an output sequence of words to form a caption sentence. Data preprocessing, model construction, and model training are discussed. Caption correctness is evaluated using 2-gram BLEU scores across the different splits of the dataset. Specific examples of output captions were shown to demonstrate model generality over (...) the video temporal dimension. Predicted captions were shown to generalize over video action, even in instances where the video scene changed dramatically. Model architecture changes are discussed to improve sentence grammar and correctness. (shrink)

This work demonstrates the implementation and use of an encoder-decoder model to perform a many-to-many mapping of video data to text captions. The many-to-many mapping occurs via an input temporal sequence of video frames to an output sequence of words to form a caption sentence. Data preprocessing, model construction, and model training are discussed. Caption correctness is evaluated using 2-gram BLEU scores across the different splits of the dataset. Specific examples of output captions were shown to demonstrate model generality over (...) the video temporal dimension. Predicted captions were shown to generalize over video action, even in instances where the video scene changed dramatically. Model architecture changes are discussed to improve sentence grammar and correctness. (shrink)

Speech Emotion Detection (SED) refers to the identification of human emotions based on speech signals. The goal of this research is to design and implement a system that can accurately classify emotions from speech using machine learning techniques. The system can be applied in various fields such as healthcare, customer service, human-computer interaction, and mental health monitoring. The paper discusses the various stages of building such a system, from collecting and preprocessing audio data to selecting machine learning models and evaluating (...) the performance of the system. A combination of feature extraction techniques and machine learning algorithms, such as Support Vector Machines (SVM) and deep learning models like Long Short-Term Memory (LSTM) networks, is employed to achieve high accuracy. (shrink)

LSTM has the advantage of handling inconsistent gradients, which was a typical problem found in conventional RNN by its cyclic rescan of the sequence to minimize the gradient errors. This is especially significant, that cryptocurrency has a highly volatile fluctuation of its market trend.

Image captioning has seen remarkable advancements with the integration of deep learning techniques, notably Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks, for generating descriptive captions for images. Despite these improvements, capturing intricate details and context remains a challenge. This project introduces an enhanced image captioning model that integrates transformers with an attention mechanism to address these limitations. By leveraging CNNs for feature extraction and LSTMs for sequence generation, while utilizing transformers to apply sophisticated attention to significant (...) image regions, the proposed model aims to generate more contextually rich and coherent captions. Experimental results indicate that incorporating transformers with attention mechanisms leads to a significant enhancement in caption accuracy and descriptiveness, surpassing traditional CNN-LSTM models. This advancement is particularly beneficial in various applications, including assistive technologies for the visually impaired, content-based image retrieval systems, automatic image annotation for digital asset management, and improved human-computer interaction. This approach represents a substantial step forward in achieving more precise and detailed image captioning, with potential impacts across numerous fields. (shrink)

Cloud computing has transformed data management by providing scalable and on-demand services, but its open and shared infrastructure makes it highly vulnerable to sophisticated cyber threats. Traditional Intrusion Detection Systems (IDS) struggle with dynamic and large-scale cloud environments due to high false positives, limited adaptability, and computational overhead. To address these challenges, this paper proposes an AI-driven Intrusion Detection System (AI-IDS) that leverages deep learning models, including Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks, to analyze network (...) traffic, detect anomalies, and identify advanced cyber threats with high accuracy. Unlike rule-based IDS, which rely on static signatures, our model continuously learns from real-time data, improving detection rates while reducing false alarms. The proposed AI-IDS is optimized for cloud infrastructure, ensuring low-latency threat detection, minimal computational burden, and real-time adaptability to emerging attack patterns. Experimental validation on benchmark datasets demonstrates significant improvements in accuracy, precision, and scalability, making AI-powered IDS a crucial innovation in cloud security. (shrink)

Bitcoin, as a decentralized digital currency, has undergone extreme price fluctuations over the years. Predicting its future price presents a significant challenge due to its volatile nature and susceptibility to various external factors, including market sentiment, regulations, and technological developments. This research aims to build an advanced forecasting model to predict Bitcoin’s price movements accurately. We leverage historical price data and apply cutting-edge machine learning techniques, including Long Short-Term Memory (LSTM) networks and Gradient Boosting Machines (GBM). By comparing these (...) methods with existing models, we evaluate their performance in predicting Bitcoin’s future prices, which is vital for traders, investors, and financial institutions in making informed decisions. (shrink)

Efficient rake scheduling and demand forecasting in railway operations are essential to address the complexities of passenger demand, minimize delays, and enhance utilization. This project uses advanced machine learning methods, specifically LSTM (Long Short-Term Memory) networks and GBM (Gradient Boosting Machine), to predict demand and optimize rake scheduling dynamically. Integrating a user-friendly web interface allows realtime data monitoring, enabling railway operators to make informed decisions. By leveraging real-time data sources, including rake movement, schedules, weather, and traffic conditions, this project (...) aims to improve operational efficiency and responsiveness. Keywords. Railway Scheduling, Rake Optimization, Demand Forecasting, LSTM, GBM, Machine Learning, RealTime Data. (shrink)

Cloud computing has revolutionized the storage and management of data by offering scalable, cost-effective, and flexible solutions. However, it also introduces significant security concerns, particularly related to data leakage, where sensitive information is exposed to unauthorized entities. Data leakage can result in substantial financial losses, reputational damage, and legal complications. This paper proposes a deep learning-based framework for detecting data leakage in cloud environments. By leveraging advanced neural network architectures, such as Long Short- Term Memory (LSTM) and Convolutional Neural (...) Networks (CNNs), the model detects abnormal data access patterns that may indicate leakage. The system operates in real-time, continuously monitoring data interactions between users and the cloud. A large dataset containing normal and abnormal access logs is used to train and validate the model, ensuring it can effectively differentiate between legitimate and malicious activity. The performance of the model is evaluated using metrics such as accuracy, precision, recall, and F1-score, with the system achieving over 96% accuracy in identifying potential data leaks. Furthermore, the proposed solution is designed to be scalable and adaptable, making it suitable for dynamic cloud environments with evolving threats. Future enhancements to the system include integrating multi- cloud support and refining the model’s ability to detect sophisticated insider threats. This research highlights the importance of leveraging deep learning for real-time, proactive cloud security. (shrink)

Nghiên cứu đánh giá hiệu quả của các mô hình học máy trong việc dự đoán biến động của chỉ số VNIndex. Kết quả nghiên cứu cho thấy, phương pháp mạng tích chập thời gian (Temporal Convolutional Networks - TCN) và mạng bộ nhớ dài ngắn (Long Short - Term Memory - LSTM) có khả năng dự báo biến động chỉ số VNIndex với độ chính xác cao, trong đó LSTM thể hiện có hiệu quả dự báo tốt (...) hơn. Phát hiện từ nghiên cứu này không chỉ góp phần vào lý thuyết dự báo tài chính mà còn cung cấp công cụ hữu ích cho các nhà đầu tư và các nhà quản lý quỹ trong việc ra quyết định đầu tư trên thị trường chứng khoán Việt Nam. [This research evaluates the effectiveness of machine learning models in predicting VNIndex fluctuations. Research results show that the method of temporal convolutional networks (TCN) and long short–term memory networks (LSTM) are capable of predicting VNIndex fluctuations with high accuracy, in which LSTM shows better forecasting performance. Findings from this study not only contribute to financial forecasting theory but also provide useful tools for investors and fund managers in making investment decisions in the Vietnamese stock market.]. (shrink)

A : In the rapidly evolving financial markets, accurately predicting stock prices is crucial for investors seeking to optimize their portfolios and mitigate risks. This project leverages machine learning techniques to develop a predictive model for stock price forecasting. We utilize historical stock price data, along with relevant economic indicators and market sentiment, to construct a robust dataset. Key methodologies include time series analysis, regression models, and advanced machine learning algorithms, such as Long Short-Term Memory (LSTM) networks, which excel (...) at capturing temporal dependencies in sequential data. The project involves comprehensive data preprocessing and feature engineering to enhance model performance. Various models are trained and evaluated using metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) to ensure accuracy and reliability. The results demonstrate the potential of machine learning approaches in financial predictions, revealing patterns and insights that can inform investment strategies. This research aims to contribute to the field of predictive analytics in finance, offering a framework for future studies and practical applications in stock market forecasting. cognitive radio(CR) is a transceiver which automatically detects available channels in wireless spectrum and accordingly changes its transmission or reception parameters. In this paper, it proposes an algorithm for the energy-efficient and spectrumaware communications requirements in CR network. It enables each node to determine and regulate its transmission strategy to provide minimum energy consumption without sacrificing end-to-end delay performance and also maximizes overall spectrum utilization. Spectrum sensing is one of the essential parameter to be considered in CR networks. Therefore, the security aspect of spectrum sensing should be addressed well. Using a Trust-Worthy algorithm, it improves the trustworthiness of the Spectrum sensing in CR-Networks. It implemented using Network Simulator-2. (shrink)

Three decades ago, William Ramsey, Steven Stich & Joseph Garon put forward an argument in favor of the following conditional: if connectionist models that implement parallelly distributed processing represent faithfully human cognitive processing, eliminativism about propositional attitudes is true. The corollary of their argument (if it proves to be sound) is that there is no place for folk psychology in contemporary cognitive science. This understanding of connectionism as a hypothesis about cognitive architecture compatible with eliminativism is also endorsed by Paul (...) Churchland, a radical opponent of folk psychology and a prominent supporter of eliminative materialism. I aim to examine whether current connectionist models based on long-short term memory (LSTM) neural networks can back up these arguments in favor of eliminativism. Nonetheless, I will rather put my faith in the eliminativism of the limited domain. This position amounts to the claim that even though connectionist cognitive science has no need whatsoever for folk psychology qua theory, this does not entail the illegitimacy of folk psychology per se in other scientific domains, most notably in humanities, but only on condition that one regards folk psychology as mere heuristics. (shrink)

Cloud computing has revolutionized the storage and management of data by offering scalable, cost-effective, and flexible solutions. However, it also introduces significant security concerns, particularly related to data leakage, where sensitive information is exposed to unauthorized entities. Data leakage can result in substantial financial losses, reputational damage, and legal complications. This paper proposes a deep learning-based framework for detecting data leakage in cloud environments. By leveraging advanced neural network architectures, such as Long Short-Term Memory (LSTM) and Convolutional Neural Networks (...) (CNNs), the model detects abnormal data access patterns that may indicate leakage. The system operates in real-time, continuously monitoring data interactions between users and the cloud. A large dataset containing normal and abnormal access logs is used to train and validate the model, ensuring it can effectively differentiate between legitimate and malicious activity. The performance of the model is evaluated using metrics such as accuracy, precision, recall, and F1-score, with the system achieving over 96% accuracy in identifying potential data leaks. Furthermore, the proposed solution is designed to be scalable and adaptable, making it suitable for dynamic cloud environments with evolving threats. Future enhancements to the system include integrating multi-cloud support and refining the model’s ability to detect sophisticated insider threats. This research highlights the importance of leveraging deep learning for real-time, proactive cloud security. (shrink)

This paper introduces a novel platform designed to utilize advanced machine learning techniques for assessing individuals' aptitude in source analysis. In today's economic landscape, the ability to accurately analyze stocks and make well-informed predictions is crucial for successful investments. The study centers on the development of an interactive platform that assesses users' investment forecasts by leveraging historical financial data and balance sheets of companies. Employing sophisticated machine learning algorithms, including Long Short-Term Memory (LSTM) and linear regression, the platform encourages (...) users to proficiently analyze and forecast future business and market values. Subsequently, the platform evaluates users' predictions against actual business data to gauge accuracy. To enhance the estimation process, the platform incorporates interactive chatbot capabilities aimed at assisting users. The objective of this research is to narrow the divide between theoretical financial knowledge and strategic investment analysis, empowering users to enhance their skills and make well-informed investment decisions. (shrink)

Cloud computing has transformed data management by providing scalable and on-demand services, but its open and shared infrastructure makes it highly vulnerable to sophisticated cyber threats. Traditional Intrusion Detection Systems (IDS) struggle with dynamic and large-scale cloud environments due to high false positives, limited adaptability, and computational overhead. To address these challenges, this paper proposes an AI-driven Intrusion Detection System (AI-IDS) that leverages deep learning models, including Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks, to analyze network (...) traffic, detect anomalies, and identify advanced cyber threats with high accuracy. Unlike rule-based IDS, which rely on static signatures, our model continuously learns from real-time data, improving detection rates while reducing false alarms. The proposed AI-IDS is optimized for cloud infrastructure, ensuring low-latency threat detection, minimal computational burden, and real-time adaptability to emerging attack patterns. Experimental validation on benchmark datasets demonstrates significant improvements in accuracy, precision, and scalability, making AI-powered IDS a crucial innovation in cloud security. (shrink)

We investigate the application of Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and a hybrid CNN-LSTM model for forecasting air pollution levels based on historical data. Our experimental setup uses real-world air quality datasets from multiple regions, containing measurements of pollutants like PM2.5, PM10, CO, NO2, and SO2, alongside meteorological data such as temperature, humidity, and wind speed. The models are trained, validated, and tested using a split dataset, and their accuracy is evaluated using performance metrics (...) like Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared. (shrink)

We investigate the application of Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and a hybrid CNN-LSTM model for forecasting air pollution levels based on historical data. Our experimental setup uses real-world air quality datasets from multiple regions, containing measurements of pollutants like PM2.5, PM10, CO, NO2, and SO2, alongside meteorological data such as temperature, humidity, and wind speed. The models are trained, validated, and tested using a split dataset, and their accuracy is evaluated using performance metrics (...) like Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared. (shrink)

Speech Emotion Recognition (SER) is an essential component in human-computer interaction, enabling systems to understand and respond to human emotions. Traditional emotion recognition methods often rely on handcrafted features, which can be limited in capturing the full complexity of emotional cues. In contrast, deep learning approaches, particularly convolutional neural networks (CNNs), recurrent neural networks (RNNs), and long short-term memory (LSTM) networks, offer more robust solutions by automatically learning hierarchical features from raw audio data. This paper reviews recent advancements in (...) deep learning-based speech emotion recognition, discusses the various architectures used, and evaluates the challenges in real-world applications. We focus on the application of deep learning models to enhance the accuracy and robustness of SER, particularly in noisy environments. The study also discusses future directions for research, including multimodal emotion recognition and transfer learning to address challenges such as small datasets and cross-domain applications. (shrink)

Phishing is one of the most effective ways in which cybercriminals get sensitive details such as credentials for online banking, digital wallets, state secrets, and many more from potential victims. They do this by spamming users with malicious URLs with the sole purpose of tricking them into divulging sensitive information which is later used for various cybercrimes. In this research, we did a comprehensive review of current state-of-the-art machine learning and deep learning phishing detection techniques to expose their vulnerabilities and (...) future research direction. For better analysis and observation, we split machine learning techniques into Bayesian, non-Bayesian, and deep learning. We reviewed the most recent advances in Bayesian and non-Bayesian-based classifiers before exploiting their cor- responding weaknesses to indicate future research direction. While exploiting weaknesses in both Bayesian and non-Bayesian classifiers, we also compared each performance with a deep learning classifier. For a proper review of deep learning-based classifiers, we looked at Recurrent Neural Networks (RNN), Convolutional Neural Networks (CNN), and Long Short Term Memory Networks (LSTMs). We did an empirical analysis to evaluate the performance of each classifier along with many of the proposed state-of-the-art anti-phishing techniques to identify future research directions, we also made a series of proposals on how the performance of the under-performing algorithm can improved in addition to a two-stage prediction model. (shrink)