Search

Phishing websites pose a significant threat to online security, necessitating robust detection mechanisms to safeguard users' sensitive information. This study explores the efficacy of various deep learning architectures for phishing website detection. Initially, traditional sequential models, including recurrent neural networks (RNN), long short-term memory (LSTM), and gated recurrent unit (GRU), achieve accuracies of 95%, 96%, and 96.5%, respectively, on a curated dataset. Building upon these results, hybrid architectures that combine the strengths of traditional sequential models with state-of-the-art language representation models, bidirectional encoder representations from transformers (BERT) and XLNet, are investigated. Combinations such as RNN with BERT, BERT with LSTM, BERT with GRU, RNN with XLNet, XLNet with LSTM, and XLNet with GRU are evaluated. Through experimentation, accuracies of 94.5%, 96.5%, 96.1%, 95.7%, 97.4%, and 97%, respectively, are achieved, demonstrating the effectiveness of hybrid deep learning architectures in enhancing phishing detection performance. These findings contribute to advancing the state-of-the-art in cybersecurity practices and underscore the importance of leveraging diverse model types to combat online threats effectively. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Intelligent Transportation Systems & Applications is the property of Journal of Intelligent Transportation Systems & Applications and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Credit card fraud detection is a critical challenge in the financial industry, with substantial economic implications. Conventional machine learning (ML) techniques often fail to adapt to evolving fraud patterns and underperform with imbalanced datasets. This study proposes a hybrid deep learning framework that integrates Generative Adversarial Networks (GANs) with Recurrent Neural Networks (RNNs) to enhance fraud detection capabilities. The GAN component generates realistic synthetic fraudulent transactions, addressing data imbalance and enhancing the training set. The discriminator, implemented using various DL architectures, including Simple RNN, Long Short-Term Memory (LSTM) networks, and Gated Recurrent Units (GRUs), is trained to distinguish between real and synthetic transactions and further fine-tuned to classify transactions as fraudulent or legitimate. Experimental results demonstrate significant improvements over traditional methods, with the GAN-GRU model achieving a sensitivity of 0.992 and specificity of 1.000 on the European credit card dataset. This work highlights the potential of GANs combined with deep learning architectures to provide a more effective and adaptable solution for credit card fraud detection. [ABSTRACT FROM AUTHOR]

Recurrent neural networks (RNNs) have significantly advanced the field of machine learning (ML) by enabling the effective processing of sequential data. This paper provides a comprehensive review of RNNs and their applications, highlighting advancements in architectures, such as long short-term memory (LSTM) networks, gated recurrent units (GRUs), bidirectional LSTM (BiLSTM), echo state networks (ESNs), peephole LSTM, and stacked LSTM. The study examines the application of RNNs to different domains, including natural language processing (NLP), speech recognition, time series forecasting, autonomous vehicles, and anomaly detection. Additionally, the study discusses recent innovations, such as the integration of attention mechanisms and the development of hybrid models that combine RNNs with convolutional neural networks (CNNs) and transformer architectures. This review aims to provide ML researchers and practitioners with a comprehensive overview of the current state and future directions of RNN research. [ABSTRACT FROM AUTHOR]

Autism spectrum disorder (ASD) is a complex, severe disorder related to brain development. It impairs patient language communication and social behaviors. In recent years, ASD researches have focused on a single-modal neuroimaging data, neglecting the complementarity between multi-modal data. This omission may lead to poor classification. Therefore, it is important to study multi-modal data of ASD for revealing its pathogenesis. Furthermore, recurrent neural network (RNN) and gated recurrent unit (GRU) are effective for sequence data processing. In this paper, we introduce a novel framework for a Multi-Kernel Learning Fusion algorithm based on RNN and GRU (MKLF-RAG). The framework utilizes RNN and GRU to provide feature selection for data of different modalities. Then these features are fused by MKLF algorithm to detect the pathological mechanisms of ASD and extract the most relevant the Regions of Interest (ROIs) for the disease. The MKLF-RAG proposed in this paper has been tested in a variety of experiments with the Autism Brain Imaging Data Exchange (ABIDE) database. Experimental findings indicate that our framework notably enhances the classification accuracy for ASD. Compared with other methods, MKLF-RAG demonstrates superior efficacy across multiple evaluation metrics and could provide valuable insights into the early diagnosis of ASD. [ABSTRACT FROM AUTHOR]

For investors and financial analysts to make informed decisions, having precise forecasts of gold prices is crucial. This study examined the effectiveness of various time series models in predicting gold prices in Indian Rupeea variety of models, ranging from linear models like Auto Regressive Integrated Moving Average (ARIMA) and Simple Linear Regression, to more complex nonlinear models such as Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU) were utilised. In the study, a statistical technique was used to analyse the data collected over a twenty-year period, from January 2001 to December 2019. RMSE and MAPE were used to evaluate the models' performance. The Simple Linear Regression model achieved an RMSE value of 834.67 and a MAPE value of 2.22%, demonstrating its superior performance compared to the other models. The LSTM and GRU models achieved RMSE values of 1160.5 and 1214.8, respectively, suggesting comparable levels of performance. The MAPE values of the LSTM model and the GRU model differed by 2.96%, with the latter being 2.83%. On the other hand, the ARIMA model had a MAPE value of 22.9% and an RMSE value of 7121.1, which was noticeably lower than the previous model. Moreover, the results showed that both LSTM and GRU have the ability to capture non-linear correlations in the fluctuations of gold prices.

Chaotic systems are identified as nonlinear, deterministic dynamic systems that are exhibit sensitive to initial values. Some chaotic equations modeled from daily events involve time information and generate chaotic time series that are sequential data. Through successful prediction studies conducted on the generated chaotic time series, forecasts can be made about events displaying unpredictable behavior in nature, which have not yet been modeled. This enables preparation for both favorable and unfavorable situations that may arise. In this study, chaotic time series were generated using Lorenz, Chen, and Rikitake multivariate chaotic systems. To enhance prediction accuracy on the generated data, GRU, LSTM and RNN models were trained with different hyperparameters. Subsequently, comprehensive test studies were conducted to evaluate their performance. Predictions were calculated using evaluation metrics, including MSE, RMSE, MAE, MAPE, and R2. In the experimental study, each chaotic system was trained with different hyperparameter combinations on six network models. The experimental results indicate that the utilized models exhibited greater success in predicting chaotic time series compared to some other models in the literature. [ABSTRACT FROM AUTHOR]

The article compares cryptocurrency price forecasting methods by evaluating Auto-Regressive Integrated Moving Average (ARIMA) and Seasonal ARIMA (SARIMA) against Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks. It discusses model effectiveness, optimization for various cryptocurrencies, and evaluates performance using metrics like Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE).

AbstractCrystallization deposition within micro-scale heat exchangers poses significant challenges to their efficiency and functionality, stemming from the accumulation of crystalline residues on heat transfer surfaces. This study employs advanced machine learning methodologies, including GRU, LSTM, RNN, and CNN models, to explore the underlying factors influencing micro heat exchanger performance. Through meticulous analysis of key parameters such as Reynolds number, sedimentation coefficient, flow rate, and channel dimensions, the study aims to delineate the foundational factors shaping heat exchanger performance at the microscopic level. Results reveal the exceptional accuracy of CNN model in forecasting experimental outcomes, surpassing 99% accuracy and demonstrating superior performance compared to traditional numerical methods. Temperature emerges as a pivotal determinant, profoundly influencing crystallization dynamics, with its intricate interplay with solute solubility elucidated through rigorous analysis. Furthermore, comparative assessment of training times highlights the CNN model’s efficiency, attributed to its specialized architecture suited for spatial data processing. This study provides valuable insights into the impact of crystallization deposition on microscale heat exchangers, showcasing the transformative potential of machine learning in optimizing heat exchanger performance and addressing operational challenges in industrial applications. [ABSTRACT FROM AUTHOR]

Several advanced driver assistance systems (ADASs) control a vehicle in the longitudinal direction. However, an ADAS that controls the vehicle in the lateral direction is uncommon since it requires the accurate lateral position prediction of the target vehicle because of the small safety margin in this direction. To reduce this problem, we suggest a data-based vehicle trajectory prediction model that mimics the human ability to predict the trajectory. The proposed model focuses on the lane-change maneuver because it is the most frequent and hard to predict from the road geometry, unlike other lateral maneuvers. The model is composed of four models to acquire interpretable outcomes. The first model predicts the longitudinal trajectory. The second and third models predict the lane-change maneuver and the time to lane change, and the last model predicts the lateral trajectory. These models are based on a recurrent neural network to consider the sequential characteristics of the input data. To train the proposed model, we generated a dataset that includes a vehicle's lateral dynamics information using the NGSIM I-80 dataset. To validate the proposed model, a test set in the dataset is used. The proposed model shows better accuracy than baseline methods based on vehicle kinematics. [ABSTRACT FROM AUTHOR]

For investors and financial analysts to make informed decisions, having precise forecasts of gold prices is crucial. This study examined the effectiveness of various time series models in predicting gold prices in Indian Rupeea variety of models, ranging from linear models like Auto Regressive Integrated Moving Average (ARIMA) and Simple Linear Regression, to more complex nonlinear models such as Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU) were utilised. In the study, a statistical technique was used to analyse the data collected over a twenty-year period, from January 2001 to December 2019. RMSE and MAPE were used to evaluate the models' performance. The Simple Linear Regression model achieved an RMSE value of 834.67 and a MAPE value of 2.22%, demonstrating its superior performance compared to the other models. The LSTM and GRU models achieved RMSE values of 1160.5 and 1214.8, respectively, suggesting comparable levels of performance. The MAPE values of the LSTM model and the GRU model differed by 2.96%, with the latter being 2.83%. On the other hand, the ARIMA model had a MAPE value of 22.9% and an RMSE value of 7121.1, which was noticeably lower than the previous model. Moreover, the results showed that both LSTM and GRU have the ability to capture nonlinear correlations in the fluctuations of gold prices. [ABSTRACT FROM AUTHOR]

The field of cybersecurity constantly evolves as attackers develop new methods and technologies. Defending against cyberattacks involves a combination of robust security measures, regular updates, user education, and the use of advanced technologies, such as intrusion detection systems and artificial intelligence, to find out the threats in realtime. IDS are designed to identify and address any unauthorized actions or potential security threats within a computer network or system. A hybrid intrusion detection system (IDS) combines many detection techniques and strategies from different IDS types into a single, coherent solution. Combining the benefits of each approach should result in more comprehensive and effective intrusion detection. This paper outlines a proposed anomaly intrusion detection system (AIDS) framework that leverages a hybrid of deep learning strategies. It incorporates Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models, which were developed using XGBoost, and their efficacy was assessed with the NSL-KDD dataset. The evaluation of the suggested model focused on its accuracy, detection capabilities, and the rate of false positives. The outcomes of this research are noteworthy within the cybersecurity field. In this paper, a framework of an Anomaly IDS is proposed. The purpose of an anomaly IDS, or AIDS, is to spot odd behavior on a network or system that might point to a security breach or malevolent attempt to hack it. Anomalybased IDSs concentrate on finding departures from accepted typical behavior, in contrast to signature-based detection systems, which depend on a predefined database of known attack patterns. [ABSTRACT FROM AUTHOR]

Deep Learning (DL) techniques have been widely used in different Natural Language Processing (NLP) tasks. Parts of Speech (PoS) tagging is one where a wide variety of DL techniques have been experimented with across the languages. Here in the present work, Recurrent Neural Network (RNN) and Gated Recurrent Unit (GRU) based Parts of Speech taggers have been trained and modelled for Assamese, an Indo Aryan family language. Universal Parts of Speech (UPoS) tag set of 17 tags were used for the experiment. A dataset of 30000 sequences has been used for the work, which is originally a BIS tag set tagged dataset, and customized through conversion from BIS tagged sequences to UPoS tagged sequences. RNN and GRU based systems have been configured using tensorflow platform and the performance measurement was done through accuracy, precision, recall and F1 scores. The accuracy of the RNN based system has been found to be 93.78%. Precision of 94.75 and recall of 93.28 were recorded for the RNN model. Accuracy of 94.38%, precision of 95.44 and recall of 93.7 were recorded for the GRU model. RNN and GRU models respectively yield F1 scores of 94.01 and 94.56. Although PoS tagging with other tag sets like BIS have been attempted by other researchers, UPoS tagging using DL approaches for Assamese is attempted for the first time. And this baseline work with observed accuracies of 93.78 and 94.38 for RNN and GRU respectively, shall serve as reference models for further works. [ABSTRACT FROM AUTHOR]

The Transformer model is frequently employed for speech command recognition (SCR) since it supports parallelization and has internal attention. The high learning speed of this design and the absence of sequential operation, like with recurrent neural networks, are its two greatest advantages. In this work, Transformer models and data augmentation techniques to enhance the used dataset were considered to build a system for the automatic command recognition of Arabic speech. Little information is available for developing voice recognition systems in Arabic, which is recognized as a component of numerous significant languages. This study examines how the Transformer network is affected by hyperparameters to address two questions: Which hyperparameter is crucial for both task effectiveness and training efficiency? According to the results of our tests, it was found that using a Transformer with hyperparameter optimization helped the Arabic SCR system operate better. [ABSTRACT FROM AUTHOR]

This paper aims to find that predictive performance is better when earning management predictions using deep learning technology, including litigation case information filed with companies. Unlike previous accounting forensics studies, this study designed a basic earning management prediction model using related financial variables by referring to various previous studies on earning management in accounting. To present more objective and reliable verification results, this study used two measures of accrual earning management (AEM) as proxy variables for corporate earning management. It used four deep-learning classifiers: RNN, GRU, LSTM, and Transformers. In addition, before predicting earning management through deep learning analysis, regression analysis proved that the number and amount of litigation cases filed against the company significantly reduced accrual earning management. In this study, the number of litigation cases filed with the company and the amount of litigation were used as information on the company’s litigation cases. The main findings of this paper are as follows. First, as expected, when predicting the earning management level, including the company’s litigation case information, accuracy, recall, precision, AUC (area under the ROC curve), and F1 score, all showed high overall predictive performance. Second, overall predictive accuracy was higher when predicting the level of earning management, including the number of litigation cases, than when predicting the level of earning management, including the number of litigation cases. Overall predictive performance was the highest when predicting the earning management level, including the number and amount of litigation cases. Third, among the four classifiers of deep learning, Transformers’ predictive performance was the best, followed by LSTM, GRU, and RNN.

This paper presents a novel approach combining Gated Recurrent Unit (GRU) and adaptive Proportional-Integral-Derivative (PID) algorithm employed for predicting tire slip angles (TSA). In previous studies, research has shown successful prediction of tire slip angles using Recurrent Neural Network (RNN). Research has shown that when using only RNN for prediction, the model’s accuracy is low in untrained speed conditions. For this investigation, vehicle data under constant speed double lane change (DLC) conditions and S-Turn variable speed (STVS) scenarios were generated using the CarSim software. The dataset includes variables such as tire lateral force, tire longitudinal force, tire vertical force, vehicle TSA, and time. A total of three GRU models were utilized in this research. The first GRU model is employed for primary predictions, the second GRU model is utilized to predict the discrepancies between the first model’s predictions and the actual values, and the third GRU model is employed to learn variations in the PID weight parameters, facilitating further optimization of the predictions. The training of the GRU models is conducted using the TensorFlow framework. An adaptive PID was realized by utilizing the predicted PID parameter changes from the third GRU model, with the aim of further refining the prediction errors. In theory, the output values of GRU2 can be directly subtracted from GRU1. Simulation results showed that the GRU_APID algorithm performed better compared with GRU1 minus GRU2. Finally, the predictions generated by the first GRU model are combined with the corrective values generated by the last adaptive PID to obtain the ultimate TSA prediction. Simulation results have confirmed that this approach indeed enhances the model’s performance and increases TSA prediction accuracy under untrained speed conditions.

Accurate prediction of monthly runoff is critical for effective water resource management and flood forecasting in river basins. In this study, we developed a hybrid deep learning (DL) model, Fourier transform long short-term memory (FT-LSTM), to improve the prediction accuracy of monthly discharge time series in the Brahmani river basin at Jenapur station. We compare the performance of FT-LSTM with three popular DL models: LSTM, recurrent neutral network, and gated recurrent unit, considering different lag periods (1, 3, 6, and 12). The lag period, representing the interval between the observed data points and the predicted data points, is crucial for capturing the temporal relationships and identifying patterns within the hydrological data. The results of this study show that the FT-LSTM model consistently outperforms other models across all lag periods in terms of error metrics. Furthermore, the FT-LSTM model demonstrates higher Nash–Sutcliffe efficiency and R2 values, indicating a better fit between predicted and actual runoff values. This work contributes to the growing field of hybrid DL models for hydrological forecasting. The FT-LSTM model proves effective in improving the accuracy of monthly runoff forecasts and offers a promising solution for water resource management and river basin decision-making processes.

Jumping motion recognition via video is a significant contribution because it considerably impacts intelligent applications and will be widely adopted in life. This method can be used to train future dancers using innovative technology. Challenging poses will be repeated and improved over time, reducing the strain on the instructor when performing multiple times. Dancers can also be recreated by removing features from their images. To recognize the dancers’ moves, check and correct their poses, and another important aspect is that our model can extract cognitive features for efficient evaluation and classification, and deep learning is currently one of the best ways to do this for short-form video features capabilities. In addition, evaluating the quality of the performance video, the accuracy of each dance step is a complex problem when the eyes of the judges cannot focus 100% on the dance on the stage. Moreover, dance on videos is of great interest to scientists today, as technology is increasingly developing and becoming useful to replace human beings. Based on actual conditions and needs in Vietnam. In this paper, we propose a method to replace manual evaluation, and our approach is used to evaluate dance through short videos. In addition, we conduct dance analysis through short-form videos, thereby applying techniques such as deep learning to assess and collect data from which to draw accurate conclusions. Experiments show that our assessment is relatively accurate when the accuracy and F1-score values are calculated. More than 92.38% accuracy and 91.18% F1-score, respectively. This demonstrates that our method performs well and accurately in dance evaluation analysis.

In this paper, recurrent neural networks consisting of GRU and LSTM architectures are used to extract meaningful insights, characteristics, and specific patterns from previously observed, equally spaced, stock market data. The long-term dependency of nonlinear time series data can be learned using GRU and LSTM. In the first phase, the sliding window technique is used to analyse the daywise (i) open, (ii) high, (iii) low, and (iv) closing values of various stocks on the stock market to forecast the future. Performance comparisons show that the proposed GRU and LSTM networks outperform the existing models in terms of prediction accuracy. Multiple datasets were compared and the findings are: (1) the proposed model has a MAPE of 0.630, while the present model's is 1.748; (2) the MAPE of the proposed model is 0.6243, while that of the existing model is 1.92; (3) the recommended model has a MAPE of 0.7924, while the existing model's is 0.8587; (4) the recommended model's MAPE is 1.191, but the current model's MAPE is 2.99. In the second phase of the process, SMA, EMA, RSI, MACD, and ADX are chosen from among the many technical indicators and used in conjunction with OHLC to further optimise the models. [ABSTRACT FROM AUTHOR]

Shares are a popular business investment, the development of information technology now allows everyone to buy and sell shares easily online, investment players, both retail and corporate, are trying to make predictions. The purpose of this study is to find out comparative performance of learning methods in stock price prediction. There are currently many research papers discussing stock predictions. using machine learning / deep learning / neural networks, in this research the author will compare several superior methods found in the latest paper findings, including CNN, RNN LSTM, MLP, GRU and their variants. From the 16 result relationships and patterns that occur in each variable and each variable is proven to show its respective role with its own weight, in general we will summarize the conclusions in chapter V below, but in each analysis there are secondary conclusions that we can get in detail. The variable that has the most significant effect on RMSE is variable B (repeatable data) compared to other variables because it has a difference in polarity that is so far between yes and no. The configuration of input timestep (history)=7 days and output timetep (prediction)=1 day is best for the average model in general. [ABSTRACT FROM AUTHOR]

Accurate classification of electromyographic (EMG) signals is vital in biomedical applications. This study evaluates different architectures of recurrent neural networks for the classification of EMG signals associated with five movements of the right upper extremity. A Butterworth filter was implemented for signal preprocessing, followed by segmentation into 250 ms windows, with an overlap of 190 ms. The resulting dataset was divided into training, validation, and testing subsets. The Grey Wolf Optimization algorithm was applied to the gated recurrent unit (GRU), long short-term memory (LSTM) architectures, and bidirectional recurrent neural networks. In parallel, a performance comparison with support vector machines (SVMs) was performed. The results obtained in the first experimental phase revealed that all the RNN networks evaluated reached a 100% accuracy, standing above the 93% achieved by the SVM. Regarding classification speed, LSTM ranked as the fastest architecture, recording a time of 0.12 ms, followed by GRU with 0.134 ms. Bidirectional recurrent neural networks showed a response time of 0.2 ms, while SVM had the longest time at 2.7 ms. In the second experimental phase, a slight decrease in the accuracy of the RNN models was observed, standing at 98.46% for LSTM, 96.38% for GRU, and 97.63% for the bidirectional network. The findings of this study highlight the effectiveness and speed of recurrent neural networks in the EMG signal classification task. [ABSTRACT FROM AUTHOR]

Low-lying coastal cities, exemplified by Norfolk, Virginia, face the challenge of street flooding caused by rainfall and tides, which strain transportation and sewer systems and can lead to personal and property damage. While high-fidelity, physics-based simulations provide accurate predictions of urban pluvial flooding, their computational complexity renders them unsuitable for real-time applications. Using data from Norfolk rainfall events between 2016 and 2018, this study compares the performance of a previous surrogate model based on a random forest algorithm with two deep learning models: Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU). The comparison of deep learning to the random forest algorithm is motivated by the desire to utilize a machine learning architecture that allows for the future inclusion of common uncertainty quantification techniques and the effective integration of relevant, multi-modal features.

Agricultural price forecasting, with its distinctive characteristics, remains a captivating field of study. In countries like India, grappling with food security challenges, reliable and efficient price forecasting models are of utmost importance. This research focuses on accurate prediction of cumin prices by emphasizing the importance of time series forecasting and the adoption of deep learning models to overcome the limitations of traditional statistical approaches. Deep learning (DL) approaches including Deep Neural Network (DNN), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) were employed to forecast cumin prices for the entire year (365 days) of 2022. The models were trained and assessed using daily price data from 2002 to 2021 from Unjha market, Gujarat, India, with accuracy metrics including RMSE, MAPE, SMAPE, MASE, and MDA. The superior accuracy of the Stacked LSTM model, particularly its low RMSE, MAPE, and SMAPE scores, along with the highest MDA, marks it as a promising tool for future agricultural price forecasting. Its precision in predicting cumin prices, with a 5 % error pre-sowing and 18 % pre-harvesting, is particularly noteworthy during critical farming periods. These findings can guide farmers in aligning their production schedules with periods of high prices, aiding in economically more beneficial farming practices. Additionally, the model's predictive reliability can assist policymakers and traders in making data-driven decisions, thus playing a significant role in stabilizing market dynamics.

Medication supply and storage are essential components of the medical industry and distribution. Most medications have a predetermined expiration date. When the demand is met in large quantities that exceed the actual need, this leads to the accumulation of medicines in the stores, and this leads to the expiration of the materials. If demand is too low, this will have an impact on consumer happiness and drug marketing. Therefore, it is necessary to find a way to predict the actual quantity required for the organization's needs to avoid material spoilage and storage problems. A mathematical prediction model is required to assist any management in achieving the required availability of medicines for customers and safe storage of medicines. The research question is to design a system based on deep learning that can predict the amount of drugs required with high efficiency and accuracy based on the chronology of previous years.Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), Bidirectional LSTM, and Gated Recurrent Unit (GRU) are used to build prediction models. Those models allow for the optimization of inventory levels, thus reducing costs and potentially increasing sales. Various measures such as mean squared error (MSE), mean absolute squared error (MASE), root mean squared error (RMSE), and others are used to evaluate the prediction models. RNN model achieved the best result with MSE: 0.019 MAE: 0.102, RMSE: 0.0.

Financial time-series data are nonlinear, complex, influenced by many economic factors, and are difficult to predict. Several traditional statistical methods have been developed for financial time series modeling. However, because it is now easier to record, analyze, and transform big data into meaningful information, the use of machine learning algorithms in financial forecast development has increased in recent years. In this study, the data of EREGL stocks, which are among the stocks traded in the main metal market in the Borsa İstanbul index, are analyzed using time series methods and then modeled using ARIMA and deep models. In the developed deep learning method, the prediction performance improved with data preprocessing stages, feature extraction studies, and different time windows. For deep learning algorithms to be used in time-series studies, a framework of time delays must be used. In this study, scenarios for different time delays and performance comparisons are performed between ARIMA models and deep learning models using long-short term emory (LSTM), gated repeating unit (GRU), and recursive neural network (RNN) algorithms. Experimental studies demonstrate that the RNN algorithm has a better predi

Recurrent neural networks (RNNs) have significantly advanced the field of machine learning (ML) by enabling the effective processing of sequential data. This paper provides a comprehensive review of RNNs and their applications, highlighting advancements in architectures, such as long short-term memory (LSTM) networks, gated recurrent units (GRUs), bidirectional LSTM (BiLSTM), echo state networks (ESNs), peephole LSTM, and stacked LSTM. The study examines the application of RNNs to different domains, including natural language processing (NLP), speech recognition, time series forecasting, autonomous vehicles, and anomaly detection. Additionally, the study discusses recent innovations, such as the integration of attention mechanisms and the development of hybrid models that combine RNNs with convolutional neural networks (CNNs) and transformer architectures. This review aims to provide ML researchers and practitioners with a comprehensive overview of the current state and future directions of RNN research.