Search

In today's evolving landscape of video surveillance, our study introduces SuspAct, an innovative ensemble model designed to detect suspicious activities in real time swiftly. Leveraging advanced Long-term Recurrent Convolutional Networks (LRCN), SuspAct represents a significant advancement in intelligent surveillance technology. By combining insights from various LRCN models through the Majority Voting ensemble technique, SuspAct enhances its overall robustness, outperforming traditional surveillance methods. Through rigorous experimentation on large-scale datasets, we demonstrate SuspAct's superiority in proactive crime prevention, showcasing its potential to revolutionize security protocols and contribute substantially to public safety. Our work addresses the challenges posed by the escalating volume of video data and lays a strong foundation for future advancements in intelligent video surveillance technology. [ABSTRACT FROM AUTHOR]

In the present day, the health of the populace is significantly jeopardized by the presence of contaminated water, and the majority of the population is unaware of the distinction between safe and unsafe water consumption. Agricultural, industrial, and other human-induced activities are causing a significant decline in the availability of drinking water. Consequently, the issue of ensuring the safety of ingesting water is becoming increasingly prevalent. People should be aware of the purity of the water and the locations where it can be used in order to resolve this situation. There are numerous IoT-based system architectures that are capable of monitoring water parameters; however, the majority of these architectures do not allow for real-time water quality prediction or visualization. In order to achieve this, we suggest a wireless framework that is based on the Internet of Things (IoT). The sensors are able to capture water parameters and transmit the data to the cloud, where a machine learning (ML) model operates to classify the water quality. After that, Grafana enables us to effortlessly visualize the real-time data and predictions from any location. We employed a multi-class dataset from China for the model's construction. GridSearchCV was implemented to identify the optimal parameters for model optimization. The proposed model is a combination of the Random Forest (RF), Extreme Gradient Boosting (XGB), and Histogram Gradient Boosting (HGB) models. The accuracy of the model for the China dataset was 99.80%. To assess the robustness of the proposed model, we acquired a new dataset from the Bangladesh Water Development Board (BWDB) and used it to test the proposed model. The model's accuracy for this dataset was 99.72%. In summary, the proposed wireless IoT framework enables individuals to effortlessly monitor the purity of water and view its parameters from any location. [ABSTRACT FROM AUTHOR]

ABSTRACTThe goal of this study is to create a dynamic model that can accurately forecast the estimated time of arrival of a bus at a specific bus stop using data obtained from the global positioning system (GPS) in Addis Ababa, Ethiopia. Both operators and customers use accurate and timely bus arrival information. In turn, it helps determine the correct placement of charging stations when green-energy vehicles are introduced in the urban mass transit system. A plethora of machine learning models have been used for prediction. Additionally, we tailored an ensemble method based on the average prediction. The performances of these machine learning methods were estimated and compared using conventional measures, such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R-squared (R2) values. The results showed that the performance of the Random Forest technique used in this study was impressive, with MAE of 0.137, MSE of 0.054, RMSE of 0.233, and R2 value of 0. 999. The findings demonstrate that the proposed method is reliable for predicting the arrival times of urban and rural buses. Next, in order to position the charging points for Electric Vehicles (EV) along specific bus routes, self-organising map (SOM) is employed to optimise EV charging point placement locations, and the result signify better performance than the kNN and DBSCAN clustering approaches.

Internet of Things (IoT) devices are leading to advancements in innovation, efficiency, and sustainability across various industries. However, as the number of connected IoT devices increases, the risk of intrusion becomes a major concern in IoT security. To prevent intrusions, it is crucial to implement intrusion detection systems (IDSs) that can detect and prevent such attacks. IDSs are a critical component of cybersecurity infrastructure. They are designed to detect and respond to malicious activities within a network or system. Traditional IDS methods rely on predefined signatures or rules to identify known threats, but these techniques may struggle to detect novel or sophisticated attacks. The implementation of IDSs with machine learning (ML) and deep learning (DL) techniques has been proposed to improve IDSs' ability to detect attacks. This will enhance overall cybersecurity posture and resilience. However, ML and DL techniques face several issues that may impact the models' performance and effectiveness, such as overfitting and the effects of unimportant features on finding meaningful patterns. To ensure better performance and reliability of machine learning models in IDSs when dealing with new and unseen threats, the models need to be optimized. This can be done by addressing overfitting and implementing feature selection. In this paper, we propose a scheme to optimize IoT intrusion detection by using class balancing and feature selection for preprocessing. We evaluated the experiment on the UNSW-NB15 dataset and the NSL-KD dataset by implementing two different ensemble models: one using a support vector machine (SVM) with bagging and another using long short-term memory (LSTM) with stacking. The results of the performance and the confusion matrix show that the LSTM stacking with analysis of variance (ANOVA) feature selection model is a superior model for classifying network attacks. It has remarkable accuracies of 96.92 % and 99.77 % and overfitting values of 0.33 % and 0.04 % on the two datasets, respectively. The model's ROC is also shaped with a sharp bend, with AUC values of 0.9665 and 0.9971 for the UNSW-NB15 dataset and the NSL-KD dataset, respectively. [ABSTRACT FROM AUTHOR]

Detection of aerial activities, including drones and birds, has practical implications for automating bird surveys and developing radar systems for aerial object collision detection. Convolutional neural networks (CNNs) have been extensively utilized for image recognition and classification tasks, albeit prior research predominantly focuses on single-class 'drone' classification. However, a gap persists in achieving high accuracy for multi-class classification. To address the limitations of traditional CNNs, such as vanishing gradients and the necessity for numerous layers, this study introduces a novel model termed ''MobVGG.” This model combines the architectures of MobileNetV2 and VGG16 to accurately classify images as either 'bird' or 'drone'. The dataset comprises 4212 images for each category of 'bird' and 'drone'. The stringent methodology was applied for dataset preparation and model training to ensure the reliability of the findings. Comparative analysis with previous research demonstrates that the proposed MobVGG model, trained on both 'bird' and 'drone' images, achieves superior accuracy (96 %) compared to benchmark studies. Our paper targets researchers and graduate students as its primary audience.

The Internet of Things (IoT) has experienced significant growth and plays a crucial role in daily activities. However, along with its development, IoT is very vulnerable to attacks and raises concerns for users. The Intrusion Detection System (IDS) operates efficiently to detect and identify suspicious activities within the network. The primary source of attacks originates from external sources, specifi-cally from the internet attempting to transmit data to the host network. IDS can identify unknown attacks from network traffic and has become one of the most effective network security. Classification is used to distinguish between normal class and attacks in binary classification problem. As a result, there is a rise in the false positive rates and a decrease in the detection accuracy during the model's training. Based on the test results using the ensemble technique with the ensemble learning XGBoost and LightGBM algorithm, it can be concluded that both binary classification problems can be solved. The results using these ensemble learning algorithms on the ToN IoT Dataset, where binary classification has been performed by combining multiple devices into one, have demonstrated improved accuracy. Moreover, this ensemble approach ensures a more even distribution of accuracy across each device, surpassing the findings of previous research.

Abstract Internet of Things (IoT) technology has revolutionized modern industrial sectors. Moreover, IoT technology has been incorporated within several vital domains of applicability. However, security is overlooked due to the limited resources of IoT devices. Intrusion detection methods are crucial for detecting attacks and responding adequately to every IoT attack. Conspicuously, the current study outlines a two-stage procedure for the determination and identification of intrusions. In the first stage, a binary classifier termed an Extra Tree (E-Tree) is used to analyze the flow of IoT data traffic within the network. In the second stage, an Ensemble Technique (ET) comprising of E-Tree, Deep Neural Network (DNN), and Random Forest (RF) examines the invasive events that have been identified. The proposed approach is validated for performance analysis. Specifically, Bot-IoT, CICIDS2018, NSL-KDD, and IoTID20 dataset were used for an in-depth performance assessment. Experimental results showed that the suggested strategy was more effective than existing machine learning methods. Specifically, the proposed technique registered enhanced statistical measures of accuracy, normalized accuracy, recall measure, and stability.

Life is easier with web applications. However, data stealing has become a serious issue and can have unpleasant effects on human life. This is when attackers try to access sensitive users' information or acquire illegal rights. Cross-site scripting is the most common attack that steals user information. It is also known as an XSS attack and appears on OWASP's Top 10 Attacks as well. More than 70 percent of web-based applications are susceptible to XSS attacks. This paper presents a Grey Wolf optimizer based on the XGboost multilayer stacking approach for XSS attack detection. The grey wolf optimizer is also known as the GWO as feature optimization for machine learning. This is superior to FEP, PSO, and GSA algorithms as it overcomes real-world challenges. The proposed techniques categorize XSS and non-XSS with an accuracy of 99.5%, recall rate of 1, false positive rate of 0.004, precision of 98.5%, and F-degree of 99.28%. The proposed methodology also compares with several existing machine learning algorithms and ensemble classifier techniques. It shows that the proposed approach is more accurate than earlier. [ABSTRACT FROM AUTHOR]

Feature importance strategy that substantially impacts software development effort estimation (SDEE) can help lower the dimensionality of dataset size. SDEE models developed to estimate effort, time, and wealth required to accomplish a software product on a limited budget are used more frequently by project managers as decision-support tool effort estimation algorithms trained on a dataset containing essential elements to improve their estimation accuracy. Earlier research worked on creating and testing various estimation methods to get accurate. On the other hand, ensemble produces superior prediction accuracy than single approaches. Therefore, this study aims to identify, develop, and deploy an ensemble approach feasible and practical for forecasting software development activities with limited time and minimum effort. This paper proposed a collaborative system containing a multi-level ensemble approach. The first level grabs the optimal features by adopting boosting techniques that impact the decided target; this subset features forward to the second level developed by a stacked ensemble to compute the product development effort concerning lines of code (LOC) and actual. The proposed model yields high accuracy and is more accurate than distinct models. [ABSTRACT FROM AUTHOR]

Question Answering (QA) systems are increasingly essential in educational institutions, enhancing both learning and administrative processes by providing quick and accurate answers to user queries. However, existing systems often struggle with accurately classifying and responding to diverse and context-dependent questions, especially when dealing with large knowledge graphs. Predicting the domain of a question can significantly narrow down the search space within a vast knowledge graph, improving the system’s efficiency and accuracy. This study addresses this gap by developing and evaluating domain prediction models. We compare the performance of various deep learning architectures, including Bi-GRU, Bi-LSTM, GRU, and LSTM. Our results demonstrate that the 1-layer Bi-GRU model outperforms the others, achieving the highest test accuracy of 82.13%. Additionally, by employing an ensemble technique that combines models with highest performance measures from each architecture, we further enhance overall performance, achieving an accuracy of 87.14%, which demonstrates improved predictive capability. This work is significant as it provides a robust solution for improving the accuracy and relevance of QA systems in educational settings, thereby enhancing user satisfaction and operational efficiency.

The automotive industry is undoubtedly taking giant strides toward a paradigm shift. In essence, wireless network communication and artificial intelligence technologies are stimulating the gradual evolution of the autonomy of intelligent vehicles. This shift causes a divergence in vehicle architecture to become assembled with software-driven rather than mechanical-driven components, producing an integrated connected central unit that perceives and processes the surrounding environment, makes autonomous decisions, and controls the entire vehicle. The emerging vehicular network technologies, including vehicle-to-everything and in-vehicle channels, facilitate wired and wireless bidirectional communication within the vehicle and to other vehicles, infrastructure actors, and the Cloud to integrate it with its surrounding environment in real-time. Despite the promised potential benefits of intelligent vehicles, including improved mobility, driving safety, and economic and environmental gains, such increased network connectivity and complexity expose them to a vast attack surface. Researchers have identified a wide range of internal and external security threats due to connectivity and automation vulnerabilities within the vehicle-to-everything wireless network channels and the lack of core security measures, such as authentication, authorization, and encryption within the in-vehicle network. The dynamic nature of these vehicular networks and their ever-changing threat landscape originate new pressing security challenges that can cause severe safety destruction. In this paper, we propose BoostSec, a novel online security analytics solution that leverages advanced incremental ensemble learning to provide robust, rapid, and adaptive protection of vehicular networks against known and unknown attacks. We further augment the proposed solution with an agnostic interpretability analysis of the results. We conducted extensive experiments on three publicly available benchmark datasets representing vehicular environments in various contexts. The experimental evaluation proves that the proposed framework outpaces current baseline approaches and meets the challenges with remarkable performance, demonstrated by its (1) generalization covering a wide range of attacks across various vehicular network contexts; (2) real-time analysis reflected in efficient computation footprint; (3) adaptability against unseen attacks; (4) robustness against adversarial attacks; and (5) augmented interpretability analysis. [ABSTRACT FROM AUTHOR]

Malignant lymphoma, which impacts the lymphatic system, presents diverse challenges in accurate diagnosis due to its varied subtypes—chronic lymphocytic leukemia (CLL), follicular lymphoma (FL), and mantle cell lymphoma (MCL). Lymphoma is a form of cancer that begins in the lymphatic system, impacting lymphocytes, which are a specific type of white blood cell. This research addresses these challenges by proposing ensemble and non-ensemble transfer learning models employing pre-trained weights from VGG16, VGG19, DenseNet201, InceptionV3, and Xception. For the ensemble technique, this paper adopts a stack-based ensemble approach. It is a two-level classification approach and best suited for accuracy improvement. Testing on a multiclass dataset of CLL, FL, and MCL reveals exceptional diagnostic accuracy, with DenseNet201, InceptionV3, and Xception exceeding 90% accuracy. The proposed ensemble model, leveraging InceptionV3 and Xception, achieves an outstanding 99% accuracy over 300 epochs, surpassing previous prediction methods. This study demonstrates the feasibility and efficiency of the proposed approach, showcasing its potential in real-world medical applications for precise lymphoma diagnosis. [ABSTRACT FROM AUTHOR]

This study aimed to develop an ensemble machine learning (ML) model for multi-step ahead SSL modeling in the Katar catchment, Ethiopia. To do so, different ML models such as multilinear regression (MLR), Feed-forward Neural Network (FFNN), Support Vector Regression and Adaptive Neuro-Fuzzy Inference System (ANFIS) were applied for one, two and three-step ahead SSL modeling. For this, two years of daily discharge and SSL data were used for model calibration and validation. Finally, four ensemble techniques: neuro-fuzzy ensemble (NFE), neural network ensemble (NE), weighted average ensemble (WE) and simple average ensemble (SE), were developed to improve the performance of single models. The performance of the developed models was evaluated using percent bias (PBIAS), mean absolute error (MAE), root mean square error (RMSE) and Nash Sutcliffe Efficiency Coefficient (NSE). The result shows that ANFIS outperformed the other individual models with a validation phase NSE value of 0.916,0.9 and 0.88 and RMSE value of 1630.5 ton/day, 1850.6 ton/day and 2026.6 ton/day, for one, two and three steps-ahead predictions, respectively. The NFE technique improved the individual model's performance in the validation phase up to 42.17%, 49.84% and 60.66% for one, two and three-step ahead modeling. Generally, the use of ensemble techniques resulted in promising improvements in single and multi-step ahead SSL modeling. [ABSTRACT FROM AUTHOR]

Globally, breast cancer (BC) is considered a major cause of death among women. Therefore, researchers have used various machine and deep learning-based methods for its early and accurate detection using X-ray, MRI, and mammography image modalities. However, the machine learning model requires domain experts to select an optimal feature, obtains a limited accuracy, and has a high false positive rate due to handcrafting features extraction. The deep learning model overcomes these limitations, but these models require large amounts of training data and computation resources, and further improvement in the model performance is needed. To do this, we employ a novel framework called the Ensemble-based Channel and Spatial Attention Network (ECS-A-Net) to automatically classify infected regions within BC images. The proposed framework consists of two phases: in the first phase, we apply different augmentation techniques to enhance the size of the input data, while the second phase includes an ensemble technique that parallelly leverages modified SE-ResNet50 and InceptionV3 as a backbone for feature extraction, followed by Channel Attention (CA) and Spatial Attention (SA) modules in a series manner for more dominant feature selection. To further validate the ECS-A-Net, we conducted extensive experiments between several competitive state-of-the-art (SOTA) techniques over two benchmarks, including DDSM and MIAS, where the proposed model achieved 96.50% accuracy for the DDSM and 95.33% accuracy for the MIAS datasets. Additionally, the experimental results demonstrated that our network achieved a better performance using various evaluation indicators, including accuracy, sensitivity, and specificity among other methods. [ABSTRACT FROM AUTHOR]

The advancement in cancer research using high throughput technology and artificial intelligence (AI) is gaining momentum to improve disease diagnosis and targeted therapy. However, the complex and imbalanced data with high dimensionality pose significant challenges for computational approaches and multi-omics data analysis. This study focuses on predicting skin cancer and analyzing overall survival probability. We employ the Kaplan–Meier estimator and Cox proportional hazards regression model, utilizing high-throughput machine learning (ML)-based ensemble methods. Our proposed ML-based ensemble techniques are applied to a publicly available dataset from the ICGC Data Portal, specifically targeting skin cutaneous melanoma cancers (SKCM). We used eight baseline classifiers, namely, random forest (RF), decision tree (DT), gradient boosting (GB), AdaBoost, Gaussian naïve Bayes (GNB), extra tree (ET), logistic regression (LR), and light gradient boosting machine (Light GBM or LGBM). The study evaluated the performance of the proposed ensemble methods and survival analysis on SKCM. The proposed methods demonstrated promising results, outperforming other algorithms and models in terms of accuracy compared to traditional methods. Specifically, the RF classifier exhibited outstanding precision results. Additionally, four different ensemble methods (stacking, bagging, boosting, and voting) were created and trained to achieve optimal results. The performance was evaluated and interpreted using accuracy, precision, recall, F1 score, confusion matrix, and ROC curves, where the voting method achieved a promising accuracy of 99%. On the other hand, the RF classifier achieved an outstanding accuracy of 99%, which exhibits the best performance. We compared our proposed study with the existing state-of-the-art techniques and found significant improvements in several key aspects. Our approach not only demonstrated superior performance in terms of accuracy but also showcased remarkable efficiency. Thus, this research work contributes to diagnosing SKCM with high accuracy. [ABSTRACT FROM AUTHOR]

In the present day, the health of the populace is significantly jeopardized by the presence of contaminated water, and the majority of the population is unaware of the distinction between safe and unsafe water consumption. Agricultural, industrial, and other human-induced activities are causing a significant decline in the availability of drinking water. Consequently, the issue of ensuring the safety of ingesting water is becoming increasingly prevalent. People should be aware of the purity of the water and the locations where it can be used in order to resolve this situation. There are numerous IoT-based system architectures that are capable of monitoring water parameters; however, the majority of these architectures do not allow for real-time water quality prediction or visualization. In order to achieve this, we suggest a wireless framework that is based on the Internet of Things (IoT). The sensors are able to capture water parameters and transmit the data to the cloud, where a machine learning (ML) model operates to classify the water quality. After that, Grafana enables us to effortlessly visualize the real-time data and predictions from any location. We employed a multi-class dataset from China for the model’s construction. GridSearchCV was implemented to identify the optimal parameters for model optimization. The proposed model is a combination of the Random Forest (RF), Extreme Gradient Boosting (XGB), and Histogram Gradient Boosting (HGB) models. The accuracy of the model for the China dataset was 99.80%. To assess the robustness of the proposed model, we acquired a new dataset from the Bangladesh Water Development Board (BWDB) and used it to test the proposed model. The model’s accuracy for this dataset was 99.72%. In summary, the proposed wireless IoT framework enables individuals to effortlessly monitor the purity of water and view its parameters from any location.

Internet of Things (IoT) devices are leading to advancements in innovation, efficiency, and sustainability across various industries. However, as the number of connected IoT devices increases, the risk of intrusion becomes a major concern in IoT security. To prevent intrusions, it is crucial to implement intrusion detection systems (IDSs) that can detect and prevent such attacks. IDSs are a critical component of cybersecurity infrastructure. They are designed to detect and respond to malicious activities within a network or system. Traditional IDS methods rely on predefined signatures or rules to identify known threats, but these techniques may struggle to detect novel or sophisticated attacks. The implementation of IDSs with machine learning (ML) and deep learning (DL) techniques has been proposed to improve IDSs’ ability to detect attacks. This will enhance overall cybersecurity posture and resilience. However, ML and DL techniques face several issues that may impact the models’ performance and effectiveness, such as overfitting and the effects of unimportant features on finding meaningful patterns. To ensure better performance and reliability of machine learning models in IDSs when dealing with new and unseen threats, the models need to be optimized. This can be done by addressing overfitting and implementing feature selection. In this paper, we propose a scheme to optimize IoT intrusion detection by using class balancing and feature selection for preprocessing. We evaluated the experiment on the UNSW-NB15 dataset and the NSL-KD dataset by implementing two different ensemble models: one using a support vector machine (SVM) with bagging and another using long short-term memory (LSTM) with stacking. The results of the performance and the confusion matrix show that the LSTM stacking with analysis of variance (ANOVA) feature selection model is a superior model for classifying network attacks. It has remarkable accuracies of 96.92% and 99.77% and overfitting values of 0.33% and 0.04% on the two datasets, respectively. The model’s ROC is also shaped with a sharp bend, with AUC values of 0.9665 and 0.9971 for the UNSW-NB15 dataset and the NSL-KD dataset, respectively.

As technologies for storing time-series data such as smartwatches and smart factories become common, we are collectively accumulating a great deal of time-series data. With the accumulation of time-series data, the importance of time-series abnormality detection technology that detects abnormal patterns such as Cyber-Intrusion Detection, Fraud Detection, Social Networks Anomaly Detection, and Industrial Anomaly Detection is emerging. In the past, time-series anomaly detection algorithms have mainly focused on processing univariate data. However, with the development of technology, time-series data has become complicated, and corresponding deep learning-based time-series anomaly detection technology has been actively developed. Currently, most industries rely on deep learning algorithms to detect time-series anomalies. In this paper, we propose an anomaly detection algorithm with an ensemble of multi-point LSTMs that can be used in three cases of time-series domains. We propose our anomaly detection model that uses three steps. The first step is a model selection step, in which a model is learned within a user-specified range, and among them, models that are most suitable are automatically selected. In the next step, a collected output vector from M LSTMs is completed by stacking ensemble techniques of the previously selected models. In the final step, anomalies are finally detected using the output vector of the second step. We conducted experiments comparing the performance of the proposed model with other state-of-the-art time-series detection deep learning models using three real-world datasets. Our method shows excellent accuracy, efficient execution time, and a good F1 score for the three datasets, though training the LSTM ensemble naturally requires more time. [ABSTRACT FROM AUTHOR]

This study evaluates the performance of the Hydrologic Engineering Center-Hydrologic Modelling System (HEC-HMS), Hydrologiska Byråns Vattenbalansavdelning (HBV), Soil and Water Assessment Tool (SWAT), feedforward neural network (FFNN), adaptive neuro-fuzzy inference system (ANFIS), support vector regression (SVR) and multilinear regression (MLR) for modelling the rainfall–runoff–sediment process in Katar catchment, Ethiopia. Afterward, neural network ensemble (NE), weighted average ensemble (WE) and simple average ensemble (SE) techniques were developed to improve the performance of single models. The performance of the models was evaluated using Nash-Sutcliffe efficiency (NSE), root mean square error (RMSE) and mean absolute error (MAE). The results show that the ANFIS model outperformed the other single models for rainfall–runoff–sediment modelling. Moreover, the integration of artificial intelligence and physically-based models resulted in improved performance, with the NE technique demonstrating better accuracy by improving individual models by 5.8–27.6% for rainfall–runoff and 3.59–37.9% for suspended sediment load modelling in the validation phase. [ABSTRACT FROM AUTHOR]

Brain tumors are aberrant cells. Medical imaging helps diagnose and classify brain tumors. MRI-based brain tumor categorization is a potential medical imaging research area. Patient's tumor sizes and features vary in brain images. Radiologists struggle to classify tumors from many photos. This research proposes an efficient Deep Learning (DL)-based tumor classification system. This research presents three CNN models for brain tumor classification. VGG16, VGG19, and SqueezeNet. Image compression reduces storage space and allows detailed analysis of brain images, which must be stored for long periods for research and medicinal purposes. For medical brain picture storage, this study uses JPEG2000. Classification is done with and without compression to determine how reduction affects classification performance. The classification models reveal that VGG16 has 98.5% accuracy, VGG19 98.80%, and SqueezeNet 98.7% without reduction. The weighted average model outperforms all other base models at a 20 K-fold value of 98.8%. [ABSTRACT FROM AUTHOR]

The main aim of this study was to develop hybrid machine learning (ML)-based ensemble modeling of the rainfall-runoff process in the Katar catchment, Ethiopia. This study used four single ML models, namely the general regression neural network (GRNN), long short-term memory neural network (LSTM), extreme learning machine (ELM) and Hammerstein-Weiner (HW) for modeling the rainfall-runoff process. Subsequently, two strategies were followed to improve the performance of the single models. In the first strategy, simple average ensemble (SAE), weighted average ensemble (WAE), Hammerstein-Weiner ensemble (HWE) and Neuro-fuzzy ensemble (NFE) were developed using the results of the single models. A hybrid Boosted Regression Tree (BRT) ensemble was developed in the second strategy to enhance the single models' modeling accuracy. The study used ten years (2008–2017) of data for calibration and validation of the developed models. The performances of the developed models were assessed using root mean square error (RMSE), percent bias (PBIAS), mean absolute error (MAE) and Nash-Sutcliffe coefficient efficiency (NSE). The results of single ML models showed that the LSTM model gave the best prediction performance with NSE = 0.933 and RMSE = 5.308 m3/s in the validation phase. For ensemble modeling, the best result was obtained by NFE increasing the performance of HW, GRNN, LSTM and ELM models by 3.35%, 13.25%, 2.57% and 19.9%, respectively. Evaluation of the hybrid BRT models showed that all the hybrid models provide reliable modeling performance with LSTM-BRT demonstrating better predictive accuracy (NSE = 0.981, RMSE = 1.999 m3/s and PBIAS = 0.75%). In general, the result of this study proved the promising influence of ensemble techniques and hybrid BRT models for rainfall-runoff modeling. [ABSTRACT FROM AUTHOR]

Developmental Dyslexia is a learning disorder often discovered in school-aged children who face difficulties while reading or spelling words even though they may have average or above-average levels of intelligence. This ultimately results in anger, frustration, low self-esteem, and other negative feelings. Early detection of Dyslexia can be highly beneficial for dyslexic children as their learning needs can be properly addressed. Researchers have used several testing techniques for early discovery where the data is collected from reading and writing tests, online games, Magnetic reasoning imaging (MRI) and Electroencephalography (EEG) scans, picture and video recording. Several Machine learning techniques have also been used in this regard recently. However, existing works did not focus on the problem of the imbalanced dataset where the percentage of dyslexic participants is much higher compared to non-dyslexic participants, which is expected to be the case for pre-screening among a random population. This paper addresses the imbalanced dataset obtained from dyslexia pre-screening tests and proposes an oversampling and ensemble-based machine learning technique for the detection of Dyslexia. Simulation results show that the proposed approach improves the detection accuracy of the minority class, i.e., dyslexic patients from 80.61% to 83.52%.

Internet of things (IOT) possess cultural, commercial and social effect in life in the future. The nodes which are participating in IOT network are basically attracted by the cyber-attack targets. Attack and identification of anomalies in IoT infrastructure is a growing problem in the IoT domain. Machine Learning Based Ensemble Intrusion Detection (MLEID) method is applied in order to resolve the drawback by minimizing malicious actions in related botnet attacks on Message Queue Telemetry Transport (MQTT) and Hyper-Text Transfer Protocol (HTTP) protocols. The proposed work has two significant contributions which are a selection of features and detection of attacks. New features are chosen from Improved Ant Colony Optimization (IACO) in the feature selection, and then the detection of attacks is carried out based on a combination of their possible properties. The IACO approach is focused on defining the attacker's important features against HTTP and MQTT. In the IACO algorithm, the constant factor is calculated against HTTP and MQTT based on the mean function for each element. Attack detection, the performance of several machine learning models are Distance Decision Tree (DDT), Adaptive Neuro-Fuzzy Inference System (ANFIS) and Mahalanobis Distance Support Vector Machine (MDSVM) were compared with predicting accurate attacks on the IoT network. The outcomes of these classifiers are combined into the ensemble model. The proposed MLEID strategy has effectively established malicious incidents. The UNSW-NB15 dataset is used to test the MLEID technique using data from simulated IoT sensors. Besides, the proposed MLEID technique has a greater detection rate and an inferior rate of false-positive compared to other conventional techniques. [ABSTRACT FROM AUTHOR]

Globally, breast cancer (BC) is considered a major cause of death among women. Therefore, researchers have used various machine and deep learning-based methods for its early and accurate detection using X-ray, MRI, and mammography image modalities. However, the machine learning model requires domain experts to select an optimal feature, obtains a limited accuracy, and has a high false positive rate due to handcrafting features extraction. The deep learning model overcomes these limitations, but these models require large amounts of training data and computation resources, and further improvement in the model performance is needed. To do this, we employ a novel framework called the Ensemble-based Channel and Spatial Attention Network (ECS-A-Net) to automatically classify infected regions within BC images. The proposed framework consists of two phases: in the first phase, we apply different augmentation techniques to enhance the size of the input data, while the second phase includes an ensemble technique that parallelly leverages modified SE-ResNet50 and InceptionV3 as a backbone for feature extraction, followed by Channel Attention (CA) and Spatial Attention (SA) modules in a series manner for more dominant feature selection. To further validate the ECS-A-Net, we conducted extensive experiments between several competitive state-of-the-art (SOTA) techniques over two benchmarks, including DDSM and MIAS, where the proposed model achieved 96.50% accuracy for the DDSM and 95.33% accuracy for the MIAS datasets. Additionally, the experimental results demonstrated that our network achieved a better performance using various evaluation indicators, including accuracy, sensitivity, and specificity among other methods.

Abstract Coral reefs are very important ecosystem which are the foundation of all life on this earth, but now they are under threat. Coral bleaching are happening now at a serious rate and the ultimate goal of conservation effort toward this issue is behaviour change. One of the most important parts of conservation effort is monitoring. However, monitoring the success of the coral bleaching campaign on behaviour change requires extensive data collection so traditional methods are not effective because they require resources that may not be met. The goal of this study is to build fast and vast automation in analyzing the stage of behaviour change. Social media data has prospect to become good alternative to be used because social media usage is currently increasing every year, including Twitter. Therefore, an automatic classification model was designed which can identify the stages of behaviour change based on the Five Doors Theory on Twitter. Five Doors Theory define 5 stages of behavior change: Desirability, Enabling Context, Can Do, Buzz, and Invitation. The data was fetched through a trusted repository, Mendeley Data, with title "An Annotated Dataset for Identifying Behaviour Change Based on Five Doors Theory Under Coral Bleaching Phenomenon on Twitter". There are 1,222 tweets with keywords related to coral bleaching that have been annotated according to the behaviour change stages. There are two proposed designs: embedding extraction which utilizes the output of each encoder layer in BERTweet and stacking ensemble which uses several BERTweet models with different hyperparameters that are ensembled using a logistic regression model. The best accuracy of 0.7796 with an f1-score of 0.7945 was obtained in the stacking ensemble design scenario. The classification model created can identify each class at the stage of behaviour change well, even though the dataset is unbalanced in its distribution. The proposed design has a performance that exceeds all baseline models and the standalone BERTweet. In conclusion, the automatic classification model create the process of monitoring the stages of behavior change run effectively and efficiently so that the success of the coral bleaching campaign can be monitored and achieved.

• Lake water quality parameter prediction by AI-driven ensemble and deep learning. • Remote sensing data combined with in-situ WQPs and water level for prediction. • Uncertainty assessment for optically active and inactive WQPs. • Reduced uncertainty by data augmentation with the TimeGAN method. • Superiority of ensemble technique in predicting WQPs in Lake Urmia. The inherent uncertainty in stochastic processes, such as lake water quality modeling, makes point predictions of water quality parameters (WQPs) insufficient for effective decision-making and management when using data-driven techniques like Artificial Intelligence (AI). This study aimed to address this issue by assessing the uncertainty associated with AI-based WQP predictions through the analysis of prediction intervals (PIs). Herein, the bootstrap method alongside AI models was utilized to estimate PIs for both optically active (Electrical Conductivity, EC) and inactive (pH, Total Dissolved Solids, TDS, Sodium Adsorption Ratio, SAR, and Total Hardness, TH) parameters at selected monitoring stations within Lake Urmia, located in the northwest region of Iran. shallow AI models—including Support Vector Regression (SVR), Random Forest (RF), and Feedforward Neural Networks (FFNN)—and deep learning (DL) methods, such as Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNN), were employed in this study. An innovative ensemble post-processing technique was specifically applied to the shallow AI models to enhance their performance. WQPs were treated as dependent variables, with independent Landsat remote sensing (RS) reflectance band data serving as input, while lake water level data was included as a secondary input to enhance model accuracy and minimize errors. Results demonstrated the superiority of the ensemble technique over individual AI and DL models. Findings indicated that DL models are less appropriate for limited observed data compared to the ensemble technique. Moreover, in terms of the model's performance for PIs, the bootstrap method provided acceptable results in estimating coverage probability and coverage width criterion (CWC). The ensemble technique consistently outperformed DL models, achieving an average of 18.53% and 27.54% lower CWC compared to LSTM and TCN models. However, the uncertainty quantity, particularly for DL models, was affected by the amount of data used. Therefore, a generative neural network model of autoencoders, known as TimeGAN (Time Generative Adversarial Network), was employed to generate synthetic data, aiming to reduce input data uncertainty. The results indicated a significant reduction in average CWC values—30.99%, 26.36%, and 23.76% for LSTM, TCN, and the ensemble technique, respectively—with deeper reductions observed for the DL models. [ABSTRACT FROM AUTHOR]

Brain tumors, a major cause of death, carry a substantial socio-economic burden. It is essential to distinguish between various types of tumors, such as gliomas, meningiomas, and pituitary tumors, using MRI data. This assists radiologists and eliminates the need for risky histology biopsies. The objective of the proposed work is to create an automated system capable of distinguishing tumor types. This system integrates a deep, shallow network with the ResNet18 model for feature extraction and machine learning methods for classification on images from the Kaggle dataset. The standalone ResNet18 model produced an accuracy of 91.42 % and to enhance the accuracy, features extracted from the pool5 layer of the ResNet18 architecture were examined by chi-square algorithm. The optimal deep features extracted using the chi-square algorithm were integrated with 17 machine learning classifier techniques comprising variants of SVM, KNN, and ensemble techniques. The proposed method predicted the classes, yielding an accuracy of 99.9 % for cubic SVM among the variants of SVM, 94 % for cosine KNN among the variants of KNN, and 99.8 % for the Adaboost algorithm only, with a maximum of the top 10 features justified using local interpretable model-agnostic (LIME) and gradient-weighted class activation mapping (Grad-CAM) visualization techniques. • ResNet-18 is used in our proposed method due to minimal number of layers compared to ResNet-50 which has 177 layers and ResNet-101which has 347 layers, thus avoiding computational complexity. • Integration of pre-trained transfer learning ResNet-18 model with 17 Machine learning classifier algorithms for the prediction and classification of classes of brain tumor. • To Enhance the accuracy, the images in the dataset are preprocessed by data augmentation approach, implementation of transfer learning approch in the ResNet-18 model and the extraction of features from pool 5 layer for the selection of the optimal feature using chi-square algorithm and classification using variants of SVM, KNN and Ensemble techniques. [ABSTRACT FROM AUTHOR]

The advancement in cancer research using high throughput technology and artificial intelligence (AI) is gaining momentum to improve disease diagnosis and targeted therapy. However, the complex and imbalanced data with high dimensionality pose significant challenges for computational approaches and multi-omics data analysis. This study focuses on predicting skin cancer and analyzing overall survival probability. We employ the Kaplan–Meier estimator and Cox proportional hazards regression model, utilizing high-throughput machine learning (ML)-based ensemble methods. Our proposed ML-based ensemble techniques are applied to a publicly available dataset from the ICGC Data Portal, specifically targeting skin cutaneous melanoma cancers (SKCM). We used eight baseline classifiers, namely, random forest (RF), decision tree (DT), gradient boosting (GB), AdaBoost, Gaussian naïve Bayes (GNB), extra tree (ET), logistic regression (LR), and light gradient boosting machine (Light GBM or LGBM). The study evaluated the performance of the proposed ensemble methods and survival analysis on SKCM. The proposed methods demonstrated promising results, outperforming other algorithms and models in terms of accuracy compared to traditional methods. Specifically, the RF classifier exhibited outstanding precision results. Additionally, four different ensemble methods (stacking, bagging, boosting, and voting) were created and trained to achieve optimal results. The performance was evaluated and interpreted using accuracy, precision, recall, F1 score, confusion matrix, and ROC curves, where the voting method achieved a promising accuracy of 99%. On the other hand, the RF classifier achieved an outstanding accuracy of 99%, which exhibits the best performance. We compared our proposed study with the existing state-of-the-art techniques and found significant improvements in several key aspects. Our approach not only demonstrated superior performance in terms of accuracy but also showcased remarkable efficiency. Thus, this research work contributes to diagnosing SKCM with high accuracy.

As technologies for storing time-series data such as smartwatches and smart factories become common, we are collectively accumulating a great deal of time-series data. With the accumulation of time-series data, the importance of time-series abnormality detection technology that detects abnormal patterns such as Cyber-Intrusion Detection, Fraud Detection, Social Networks Anomaly Detection, and Industrial Anomaly Detection is emerging. In the past, time-series anomaly detection algorithms have mainly focused on processing univariate data. However, with the development of technology, time-series data has become complicated, and corresponding deep learning-based time-series anomaly detection technology has been actively developed. Currently, most industries rely on deep learning algorithms to detect time-series anomalies. In this paper, we propose an anomaly detection algorithm with an ensemble of multi-point LSTMs that can be used in three cases of time-series domains. We propose our anomaly detection model that uses three steps. The first step is a model selection step, in which a model is learned within a user-specified range, and among them, models that are most suitable are automatically selected. In the next step, a collected output vector from M LSTMs is completed by stacking ensemble techniques of the previously selected models. In the final step, anomalies are finally detected using the output vector of the second step. We conducted experiments comparing the performance of the proposed model with other state-of-the-art time-series detection deep learning models using three real-world datasets. Our method shows excellent accuracy, efficient execution time, and a good F1 score for the three datasets, though training the LSTM ensemble naturally requires more time.

Real-world nonstationary data are usually characterized by high nonlinearity and complex patterns due to the effects of different exogenous factors that make prediction a very challenging task. An ensemble strategically combines multiple techniques and tends to be robust and more precise compared to a single intelligent algorithmic model. In this work, a dynamic particle swarm optimization-based empirical mode decomposition ensemble is proposed for nonstationary data prediction. The proposed ensemble implements an environmental change detection technique to capture concept drift occurring and the intrinsic nonlinearity in time series, hence improving prediction accuracy. The proposed ensemble technique was experimentally evaluated on electric time series datasets. The obtained results show that the proposed technique improves prediction accuracy and it outperformed several state-of-the-art techniques in several cases. For future work direction, a detailed empirical analysis of the proposed technique can be considered such as the effect of the cost of prediction errors, and the technique's search capability. [ABSTRACT FROM AUTHOR]

Many researchers are using microarray technology to examine and investigate the levels of gene expression in a specific organism, which is an emerging trend in the field of genetic research. Microarray studies have a wide range of applications in the health sector, including disease prediction and diagnostics, as well as cancer research. Due to the existence of irrelevant or duplicated data in microarray datasets, it is difficult to correctly and immediately capture possible patterns using existing algorithms. Feature selection (FS) has evolved into a critical approach for identifying and eliminating the most pertinent qualities. The enormous dimensionality of microarray datasets, on the other hand, presents a significant barrier to the majority of available FS techniques. In this research, we propose a Modified Firefly Feature Selection (MFFS) algorithm that will reduce the irrelevant attributes needed for classification and a Deep Learning Model for classifying the microarray data. The experimental outcomes show that the proposed MFFS algorithm combined with a Hybrid Deep Learning Algorithm outperforms the existing methods in terms of feature set size, accuracy, precision, recall, F-measure and AUC for a dataset with larger number of features. [ABSTRACT FROM AUTHOR]