Search

This paper presents a new soft switching (SS) multiphase interleaved boost converter (MIBC) consisting of three boost converters (BCs) that operate rapidly and efficiently to charge a lithium-ion battery for photovoltaic (PV) power system applications. This suggested converter can operate both active power switches at zero current switching (ZCS), and when switched off, it works at zero voltage switching (ZVS), which reduces switching losses and increases switching efficiency (SF). An interleaving structure is employed to decrease the conduction losses and reduce the input current ripple (ICR) and output voltage ripple (OVR) to increase the power rating. The suggested converter regulates the input and output power by modulating the phase shift in the pulse width modulation (PWM) technique using the maximum power point tracking (MPPT) algorithm via two closed feedback loops, one for a slower external loop and the other for a faster PWM control internal loop, similar to current mode control. The suggested converter has high voltage (HV) gain due to the coupled inductors (CIs), which act as a transformer. This magnetic design structure in both the steady-state and transient states also decreases the size and enhances the converter’s performance. Finally, the suggested converter was experimentally verified on a 285W prototype, and the maximum efficiency was found to be 98%, based on modeling and experimental data.

Speech enhancement (SE) aims to improve the quality and intelligibility of speech signals, particularly in the presence of noise or other distortions, to ensure reliable communication and robust speech recognition. Deep neural networks (DNNs) have shown remarkable success in SE due to their ability to learn complex patterns and representations from large amounts of data. However, they face limitations in handling long-term temporal sequences. Spiking neural networks and transformers inherently manage temporal data and capture fine-grained temporal patterns in speech signals. This paper proposes a model that integrates self-attention with spiking neural networks for speech enhancement. The proposed model employs a convolutional encoder-decoder architecture with a spiking transformer acting as a bottleneck network. The spiking self-attention mechanism in this framework represents features using spike-based queries, keys, and values. This approach enhances features by effectively capturing temporal dependencies and contextual relationships in speech signals. The spiking transformer is divided into two branches to capture comprehensive global dependencies across the temporal and spectral dimensions. The encoder-decoder incorporates a multi-scale feature extractor, which extracts features at various scales, enabling the model to build a comprehensive hierarchical representation. This representation significantly enhances the model’s ability to learn and process noisy speech, leading to excellent SE performance. Experiments are conducted using two publicly available benchmark datasets: WSJO-SI84 and VCTK+DEMAND. The proposed model demonstrated improved SE performance, showing significant progress with notable improvements of 33.69% in ESTOI, 1.05 in PESQ, and 11.36 dB in SDR over the noisy mixtures.

This study proposes a novel isolated multiport bidirectional DC/DC converter (IMBC) that combines soft-switching (SS) and resistor-capacitor-diode (RCD) snubber circuits for battery charge and discharge applications in photovoltaic (PV) power systems. The proposed SS circuit operates at zero-current (ZC) switching (ZCS) when the switches are on to reduce the switching losses. The RCD snubber circuit also suppresses the overvoltage when turning off switches. Combining the two circuits also enables a more comprehensive operating range for SS, resulting in low losses and high efficiency. A thorough study of operating systems and design considerations is shown. PSIM illustrates the effectiveness of the suggested control method. A prototype of the suggested IMBC with an output voltage of 120 V and an output power of 150 W is constructed to confirm the viability and efficiency of the suggested topology. The maximum efficacy of the suggested topology in the step-up mode for the PV arrays is about 96.4% at an input voltage of 20.5 V and output power of 150 W. In comparison, the efficiency is about 98.5% in the step-down mode for charging and discharging the battery which is considered a highly satisfactory performance.

In deep learning, various optimizers use a starting value for the learning rate. In general, this value is the one defined by default in the optimizer or chosen by the user after searching for a value giving the best performances of the studied neural network. Finding this value in the case of complex models with large-scale data, either by fine-tuning it manually or by applying a grid search algorithm requires several training sessions, which is penalizing in terms of learning time. To solve this problem, we propose an efficient optimizer using an automatic setting of the learning rate value. Learning then requires a single training session. The proposed optimizer is based on the inexact linear search of Armijo’s rule that we adapt for neural networks. We have experimented with this optimizer for several types of loss functions such as mean squared error, mean absolute error, and cross-entropy. Different types of models such as regression and classification are studied. Also, the robustness of this optimizer against the noisy labels is verified. Empirical results demonstrate that the proposed optimizer works well in practice, it outperforms other state-of-the-art optimizers that use default learning rate in terms of accuracy in addition to the time saving when searching for the optimal value of the learning rate.

Environmental, safety, and availability constraints of mineral oils have led transformer manufacturers and researchers to turn to alternative liquid dielectrics to mineral oils. Thus, many works have focused on esters and more particularly on natural esters (NEs); commonly called vegetable oils; in recent decades. Besides their advantages (biodegradability, non-toxicity, recyclability, thermal stability, and safety) compared to mineral oils, natural esters have certain disadvantages including their high viscosity and their oxidation instability. Both of these properties can be improved by adding specific additives and nanoparticles. This paper is mainly focused on review of the breakdown voltage of natural ester-based nanofluids, one of the most significant parameters when selecting insulating fluids for high voltage devices and in particular for power transformers. Firstly, we take stock of the history of the introduction of fluids/oils in the field of electrical insulation and the return of vegetable oils in the transformer industry. Then, we present the main characteristics of vegetable oils and the different products based on vegetable oils currently on the market as well as those that are currently the subject of research and plants from which they come. The methods and techniques for the preparation/elaboration of these nanofluids and their stability are then exposed. Finally, we present the different results reported in the literature on the electrical breakdown of various types of natural esters when adding nanoparticles (NPs) issued from different materials (conducting, semiconducting, and insulating), under AC, DC and lightning impulse voltages, and discuss the involved mechanisms in the breakdown of nanofluids.

The successful integration of renewable energy resources into the power grid hinges on the development of energy storage technologies that are both cost-effective and reliable. These storage technologies, capable of storing energy for durations longer than 10 hours, play a crucial role in mitigating the variability inherent in wind and solar-dominant power systems. To shed light on this matter, a transparent, least-cost macro energy model with user-defined constraints has been utilized for a case study of California. The model addresses all included technologies, solving for both hourly dispatch and installed capacities. Real-world historical demand and hourly weather data have been utilized to do this analysis. A novel approach has been introduced to assess the significance of long-duration energy storage technologies (LDS) in terms of their energy and power capacity. This method explores the contributions of pumped hydropower storage (PHS), compressed air energy storage (CAES), and power-to-gas-to-power (PGP) storage toward minimizing the overall balance of system cost. Historical electricity demand, hourly weather data, and current technology costs are used to investigate high-level implications for California’s power system options. Increasing the storage capacity of each technology from 1 to 10 hours results in 29.6%, 14.4%, and 7.5% cost reduction for PHS, CAES, and PGP cases respectively. However, in studied simulations, maximum availability (maximum) of pumped hydropower storage reduces the balance of system costs by 72.3% followed by CAES (60.6%) and PGP (48.6%) and suggests that pumped hydropower storage in combination with CAES/PGP could play an important role in California’s electricity system, provided that suitable sites can be identified and constructed at reasonable costs.

The Single-Phase Induction Motor (SPIM) has gained widespread adoption in various power applications. This article introduces a new study and analysis of SPIM, focusing on its high-frequency (HF) modeling within the conducted electromagnetic interference (EMI) frequency range of interest (150 kHz– 30MHz). This research is crucial for ensuring electromagnetic compatibility (EMC), evaluating electromagnetic interference issues, optimizing machine design, and complying with EMC standards. To identify common-mode current propagation paths accurately, a precise impedance analyzer is required to determine the common-mode impedances models of the high-frequency motor. The article proposes an HF modeling method based on a genetic algorithm (GA), which offers the advantages of high accuracy, versatility, and fast simulation time. This approach effectively reproduces models, trends, and relationships observed in real data, indicating a high level of fidelity and reliability in its predictions. Case studies validate the efficiency of the proposed HF modeling method. These precise predictions of undesired HF phenomena provide valuable insights into identifying areas where parasitic currents may propagate in the motor and understanding the impact of each motor phase on the overall behavior of the motor’s common-mode impedance.

The development of data analysis techniques and intelligent systems has had a considerable impact on education, and has seen the emergence of the field of educational data mining (EDM). The Early Warning System (EWS) has been of great use in predicting at-risk students or analyzing learners’ performance. Our project concerns the development of an early warning system that takes into account a number of socio-cultural, structural and educational factors that have a direct impact on a student’s decision to drop out of school. We have worked on an original database dedicated to this issue, which reflects our approach of seeking exhaustiveness and precision in the choice of dropout indicators. The model we built performed very well, particularly with the K-Nearest Neighbor (KNN) algorithm, with an accuracy rate of over 99.5% for the training set and over 99.3% for the test set. The results are visualized using a Django application we developed for this purpose, and we show how this can be useful for educational planning.

Disabled people are facing many difficulties communicating with others and involving in society. Modern societies have dedicated significant efforts to promote the integration of disabled individuals into their societies and services. Currently, smart healthcare systems are used to facilitate disabled people. The objective of this paper is to enable two-way communication of deaf individuals with the rest of society, thus enabling their migration from marginal elements of society to mainstream contributing elements. In the proposed system, we developed three modules; the sign recognition module (SRM) that recognizes the signs of a deaf individual, the speech recognition and synthesis module (SRSM) that processes the speech of a non-deaf individual and converts it to text, and an Avatar module (AM) to generate and perform the corresponding sign of the non-deaf speech, which were integrated into the sign translation companion system called Saudi deaf companion system (SDCS) to facilitate the communication from the deaf to the hearing and vice versa. This paper also contributes to the literature by utilizing our self-developed database, the largest Saudi Sign Language (SSL) database—the King Saud University Saudi-SSL (KSU-SSL). The proposed SDCS system performs 293 Saudi signs that are recommended by the Saudi Association for Hearing Impairment (SAHI) from 10 domains (healthcare, common, alphabets, verbs, pronouns and adverbs, numbers, days, kings, family, and regions).

This paper addresses the problem of output feedback control for high voltage direct current (HVDC) systems based on voltage source converters (VSC). Specifically, the system under study consists of two power grids and two VSC stations connected via a long HVDC cable. The primary goal of this paper is to propose a different approach for the online estimation of the grid voltage placed at a great distance from the control unit which is considered generally known, moreover the proposed technique aims to online estimation of the grid pulsation and the HVDC cable states, generally considered available. A nonlinear observer is intended to address these concerns. Only the line current measurements are used to estimate all state variables. Furthermore, utilizing the estimation offered by the proposed observer, an output feedback control strategy is formulated, using the backstepping design technique, to meet the three control objectives: 1) tight output voltage regulation of the HVDC cable; 2) ensuring a perfect power quality at both sides of the HVDC system; 3) regulation of the active power at the inverter station. The high effectiveness of the proposed output feedback controller’s performance is highlighted via a semi-experimental platform that includes a Processor In the Loop (PIL) on the embedded board eZdsp. Extra numerical simulations show that The proposed output feedback performance enjoys additional robustness properties and meets all control objectives.

The expanding use of photovoltaics (PV) as a green energy resource has been rising in these years, mostly due to the possibility of being incorporated with traditional power systems, to meet the world’s energy needs and reduce carbon emissions. However, providing green electricity from this renewable generator is frequently vulnerable to power quality (PQ) disruptions resulting from the PV’s intermittent nature and other factors associated with the electric grid, power converters, and linked loads. These disruptions need to be reduced to keep the investigated system’s PQ from deteriorating. The investigated system includes PV, DC-DC, and DC-AC converters, filter, power grid, and control schemes. If the DC-DC converter is not managed, a deviation from the maximum power point (MPP) extrapolated from the PV system will take place. In order to maximize the energy harvested from the PV system by managing the DC-DC converter, this research developed two MPP tracking (MPPT) algorithms: artificial neural networks (ANN) and cuckoo search (CS). Additionally, a design and implementation for a shunt active power filter (LCL) using genetic algorithm and GRG is provided to lower the injected total harmonic distortion (THD) and thereby enhance the PQ. To achieve the smallest size of the LCL components, the generalized reduced gradient (GRG) was the best compared to genetic algorithms GA. The results of the simulation showed that ANN performed better at tracking maximum power than CS. With the designed LCL, the THD is reduced by 99.78% compared to without a filter. To verify the simulation’s findings, a practical configuration is implemented.

This paper presents a new design of an experimental low power Wind Turbine Emulator (WTE) which has the advantage of being simple and easy to implement in practice. The proposed emulator takes into account the effect of Wind Turbine (WT) inertia as well as the nonlinearities in the system which make it suitable for all types of WTs. An efficient Maximum Power Point Tracking (MPpt) algorithm based on synergetic and backstepping nonlinear control theory has also been proposed to allow the WT to extract maximum power under a wide range of operating conditions. This is achieved by regulating the current at the input of the boost converter through these controllers. Furthermore, these nonlinear controllers considerably reduce the complexity of the design of the control scheme. Several experimental test are conducted to validate the MPPT strategy and the proposed control scheme using a dSPACE1104 board and MATLAB/Simulink environment, including different wind profiles and variable electrical load. The results obtained demonstrate the effectiveness of the designed WTE in reproducing the same mechanical behavior as the real turbine studied and the good performance achieved by the proposed MPPT control algorithm.

Federated Learning (FL) is an innovative area of machine learning that enables different clients to collaboratively generate a shared model while preserving their data privacy. In a typical FL setting, a central model is updated by aggregating the clients’ parameters of the respective artificial neural network. The aggregated parameters are then sent back to the clients. However, two main challenges are associated with the central aggregation approach. Firstly, most state-of-the-art strategies are not optimised to operate in the presence of certain types of non-iid (not independent and identically distributed) applications and datasets. Secondly, federated learning is vulnerable to various privacy and security concerns related to model inversion attacks, which can be used to access sensitive information from the training data. To address these issues, we propose a novel federated learning strategy FedNets based on ensemble learning. Instead of sharing the parameters of the clients over the network to update a single global model, our approach allows clients to have ensembles of diverse-lightweight models and collaborate by sharing ensemble members. FedNets utilises graph embedding theory to reduce the complexity of running Deep Neural Networks (DNNs) on resource-limited devices. Each Deep Neural Network (DNN) is treated as a graph, from which respective graph embeddings are generated and clustered to determine which part of the DNN should be shared with other clients. Our approach outperforms state-of-the-art FL algorithms such as Federated Averaging (Fed-Avg) and Adaptive Federated Optimisation (Fed-Yogi) in terms of accuracy; on the Federated CIFAR100 dataset (non-iid), FedNets demonstrates a remarkable 63% and 92% improvement in accuracy, respectively. Furthermore, FedNets does not compromise the client’s privacy, as it is safeguarded by the design of the method.

Modern wind power systems have recently tended to focus on achieving fast-tracking wind speeds (WSs), high maximum power point tracking (MPPT) efficacy without mechanical sensors, and high performance under uncertain WS together with an effective control system. Therefore, a sensorless MPPT method is introduced, which calculates the actual WS to save system installation costs and boost performance levels. The implemented MPPT method is based on the approximating of the 3-order polynomial to the aerodynamics torque power coefficient. In this study, three-speed control strategies (SCSs) for a grid-connected permanent magnet synchronous wind generator (PMSWG) are examined and assessed. Harris Hawks’ algorithm (HHA)-based PI controller (HHA-PIC) is used in place of (the conventional proportional-integral controller (CPIC), and adaptive fuzzy logic controller (AFLC)) as a speed controller to overcome their drawbacks. To track the generator speed to the desired speed, the HHA-PIC is used. All the CPIC, AFLC, and HHA-PIC have been carefully thought out and constructed to satisfy the speed control loop’s responsive performance. Additionally, a comparison of SCSs amongst the categories under investigation is done. The effect of HHA on the functionality of SCS is verified using MATLAB/SIMULINK. To ensure the efficacy and supremacy of the HHA-PIC over the CPIC and AFLC, a wide variety of WSs (step change, ramp, and real fluctuations) are applied. The HHA-PIC boosts system efficiency over AFLC and CPIC by 0.81% and 8.48%, respectively. Finally, it can be said that HHA is a crucial remedy for the problems with CPIC and is superior to AFLC.

In this paper, we propose a new iterative thresholding algorithm based optimizer (Itao) for deep neural networks. It is a first-order gradient-based algorithm with Tikhonov regularization for stochastic objective functions. It is fast and straightforward to implement. It acts on the parameters and their gradients, with respect to the objective function, in only one step in the backpropagation system when training a neural network. This reduces the learning time and makes it well suited for neural networks with large parameters and/or large datasets. We have experimented this algorithm on several types of loss functions such as mean squared error, mean absolute error and categorical crossentropy. Different types of models such as regression and classification are studied. The robustness of this optimizer against the noisy labels is also verified. Many of the empirical results of conducted experiments in this study show that our optimizer works well in practice. It can outperform other state-of-the-art optimizers in terms of accuracy or at least give the same results in addition to the reduction of learning time.

With increased pressures and tightening budgets within English Children’s Services in the UK, seeking more effective operational and financial management is becoming a more significant topic of discussion. In other sectors, complex data analysis methods provide the aforementioned management improvements through better understanding of current situations leading to better decision making. Currently, investment remains at a slow pace in English Local Authorities due to budget restrictions. In this paper, a potential opportunity is explored with existing publicly available data related to this area. With the help of industry experts, an Agent-Based Model is created to emulate basic Children’s Services operations and optimised to fit existing data using NSGA-III. With relatively close matches being achieved with sample authorities, this approach demonstrates promise in advancing analytics capabilities for Children’s Services and practical solutions are discussed. With this presented work, it is shown that further expansion and exploration into real-world applications is warranted.

Investigations made to evaluate the site-effect characteristics and to develop a reliable site classification scheme have received paramount importance for the urban areas planning and reliable site-specific seismic hazard assessment. This paper presents a novel non-objective and data-driven approach for preliminary seismic site-specific classification maps using machine learning (ML) based on affinity propagation (AP) along with a selected set of representative horizontal to vertical spectral ratio (HVSR) curves inside King Saud University (KSU) campus, which is among the main areas in Saudi Arabia. Besides, the proposed model aims to overcome the clustering error due to the dependency of the interpreter’s experience. Measurements of the ambient vibrations were performed to cover the entire campus area by about 307 stations. Recording at each station lasted for 20 minute length and a sample rate of 128 Hz for each station to satisfy the criteria for reliable and unambiguous HVSR results. Frequency and amplification values were used for subsequent site classification by passing messages between data points. The obtained results illustrate that the microtremor spectral ratio can be a remarkably robust tool in determining site effects. Accordingly, the proposed methodology can assist the decision-makers to set the priorities of managing land uses, estimating the earthquake losses, conducting programs for reducing the vulnerability of existing structures, enforcing building codes, planning for emergency response and long-term recovery, and designing and implementing phases of new constructions.

Due to the high availability of large-scale annotated image datasets, paramount progress has been made in deep convolutional neural networks (CNNs) for image classification tasks. CNNs enable learning highly representative and hierarchical local image features directly from data. However, the availability of annotated data, especially in the medical imaging domain, remains the biggest challenge in the field. Transfer learning can provide a promising and effective solution by transferring knowledge from generic image recognition tasks to the medical image classification. However, due to irregularities in the dataset distribution, transfer learning usually fails to provide a robust solution. Class decomposition facilitates easier to learn class boundaries of a dataset, and consequently can deal with any irregularities in the data distribution. Motivated by this challenging problem, the paper presents Decompose, Transfer, and Compose (DeTraC) approach, a novel CNN architecture based on class decomposition to improve the performance of medical image classification using transfer learning and class decomposition approach. DeTraC enables learning at the subclass level that can be more separable with a prospect to faster convergence. We validated our proposed approach with three different cohorts of chest X-ray images, histological images of human colorectal cancer, and digital mammograms. We compared DeTraC with the state-of-the-art CNN models to demonstrate its high performance in terms of accuracy, sensitivity, and specificity.

The classical radial harmonic Fourier moments (RHFMs) and the quaternion radial harmonic Fourier moments (QRHFMs) are gray-scale and color image descriptors. The radial harmonic functions with integer orders are not able to extract fine features from the input images. In this paper, the authors derived novel fractional-order radial harmonic functions in polar coordinates. The obtained functions are used to defined novel multi-channel fractional-order radial harmonic moments (FrMRHFMs) for color image description and analysis. The invariants to geometric transformations for these new moments are derived. A theoretical comparison between FrMRHFMs and QRHFMs is performed from the aspects of kernel function and the spectrum analysis. Numerical simulation is carried out to test these new moments in terms of image reconstruction capabilities, invariance to the similarity transformations, color image recognition and the CPU computational times. The obtained theoretical and numerical results clearly show that the proposed FrMRHFMs is superior to the QRHFMs and the existing fractional-order orthogonal moments.

Process mining is often used by organisations to audit their business processes and improve their services and customer relations. Indeed, process execution (or event) logs constantly generated through various information systems can be employed to derive valuable insights about business operations. Compared to traditional process mining techniques such as Petri nets and the Business Process Model and Notation (BPMN), deep learning methods such as Recurrent Neural Networks, and Long Short-Term Memory (LSTM) in particular, have proven to achieve a better performance in terms of accuracy and generalising ability when predicting next events in business processes. However, unlike the traditional network-based process mining techniques that can be used to visually present the entire discovered process, the existing deep learning-based methods for process mining lack a mechanism explaining how the predictions of next events are made. This study proposes a new approach to process mining by combining the benefits of the earlier, visually explainable graph-based methods and later, more accurate but unexplainable deep learning methods. According to the proposed approach, an LSTM model is employed first to find probabilities for each known event to appear in the process next. These probabilities are then used to generate a visually interpretable process model graph that represents the decision-making process of the LSTM model. The level of detail in this graph can be adjusted using a probability threshold, allowing to address a range of process mining tasks such as business process discovery and conformance checking. The advantages of the proposed approach over existing LSTM-based process mining methods in terms of both accuracy and explainability are demonstrated using real-world event logs.

Masses are mammographic nonpalpable signs of breast cancer. These masses could be detected using screening mammography. This paper proposed a system utilizing orthogonal moment invariants (OMIs) features for mammographic masses detection and diagnosis. In this work, three sets of OMIs features were extracted. These OMIs features are Gaussian-Hermite moments (GHMs), Gegenbauer moments (GeMs), and Legendre moments (LMs). The extracted features are fused and presented to the particle swarm optimization (PSO) algorithm for feature selection. The classification step is achieved using the support vector machine (SVM). The proposed system is evaluated using 400 regions, extracted from the DDSM dataset. The obtained results reveal the promising application of OMIs features for masses detection and identification. It shows that fusing the OMIs features produces an acceptable detection performance where the area under the receiver operating characteristics (ROC) curve is $Az=0.969\pm 0.01$ and the best performance of OMIs features is $Az = 0.856\pm 0.053$ for characterizing the malignancy of masses.

In response to the increased energy consumption in residential buildings, various efforts have been devoted to increase occupant awareness using energy feedback systems. However, it was shown that the feedback provided by these systems is not enough to inform occupant actions to reduce energy consumption. Another approach is to control energy consumption using automated energy management systems. The automatic control of appliances takes out the occupant sense of control, which is proved to be uncomfortable in many cases. This paper proposes an energy messaging intervention that keeps the control for occupants while supporting them with actionable messages. The messages inform occupants about energy waste incidents happening in their house in real time, which enables occupants to take actions to reduce their consumption. Besides, a heuristic is defined to make the intervention non-intrusive by controlling the rate and time of the messages sent to occupants. The proposed intervention is evaluated in a novel layered agent-based model. The first layer of the model generates the detailed energy consumption and realistic occupant activities. The second layer is designed to simulate the peer pressure effect on the energy consumption behavior of the individuals. The third layer is a customizable layer that simulates energy interventions. The implemented intervention in this paper is the proposed non-intrusive messaging intervention. A number of scenarios are presented in the experiments to show how the model can be used to evaluate the proposed intervention and achieve energy efficiency targets.

Early detection of emerging research trends could potentially revolutionise the way research is done. For this reason, trend analysis has become an area of paramount importance in academia and industry. This is due to the significant implications for research funding and public policy. The literature presents several emerging approaches to detecting new research trends. Most of these approaches rely mainly on citation counting. While citations have been widely used as indicators of emerging research topics, they suffer from some limitations. For instance, citations can take months to years to progress and then to reveal trends. Furthermore, they fail to dig into paper content. To overcome this problem, we introduce Leap2Trend, a novel approach to instant detection of research trends. Leap2Trend relies on temporal word embeddings (word2vec) to track the dynamics of similarities between pairs of keywords, their rankings and respective uprankings (ascents) over time. We applied Leap2Trend to two scientific corpora on different research areas, namely computer science and bioinformatics and we evaluated it against two gold standards Google Trends hits and Google Scholar citations. The obtained results reveal the effectiveness of our approach to detect trends with more than 80% accuracy and 90% precision in some cases. Such significant findings evidence the utility of our Leap2Trend approach for tracking and detecting emerging research trends instantly.

The extensive growth of digital technologies such as the Internet of Things (IoT), social media networks and forecasting systems has led to new challenges regarding computational complexity and big data mining. The classification task in such applications is not trivial due to the high volume of related data and limited time available for the task. It is particularly difficult when dealing with data streams, where each instance of data is typically processed once on its arrival (i.e. online) while the underlying data distribution often changes due to the changing environment. In this paper, we propose a novel ensemble-based framework called Replicator Dynamics & Genetic Algorithms Approach (RED-GENE) for effective data stream classification in the context of changing environment leading to concept drifts (i.e. evolution of data streams). RED-GENE employs three novel Replicator Dynamics (RD) strategies along with a Genetic Algorithm (GA) optimisation technique to flexibly adapt to different types of concept drifts when performing data stream classification tasks. The proposed framework works as follows. First, a set of random feature combinations is drawn from a given pool of features of the target data stream to create different classification types. Next, RD is used to allow the classification types achieving higher classification accuracy to grow and those with lower accuracy to shrink. A modified version of the classic GA is then employed to optimise the randomly drawn combinations of features in each classification type. The proposed framework was tested using nine data streams (including both real-world and synthetic datasets) to investigate different variations of the proposed framework and compare its performance to other state-of-the-art algorithms using immediate and delayed prequential evaluation methods. The results demonstrated that the proposed framework can provide the best accuracy on average when comparing to five other state-of-the-art algorithms.

Virtual Learning Environments provide teachers with a web-based platform to create different types of feedback. These environments usually follow the `one size fits all' approach and provide students with the same feedback. Several personalized feedback frameworks have been proposed which adapt the different types of feedback based on the student characteristics and/or the assessment question characteristics. The frameworks are intradisciplinary, neglect the characteristics of the assessment question, and either hard-code or auto-generate the types of feedback from a restricted set of solutions created by a domain expert. This paper contributes to research carried out on personalized feedback frameworks by proposing a generic novel system which is called the Ontology-based Personalized Feedback Generator (OntoPeFeGe). OntoPeFeGe addressed the aforementioned drawbacks using an ontology-a knowledge representation of the educational domain. It integrated several generation strategies and templates to traverse the ontology and auto-generate the questions and feedback. The questions have different characteristics, in particular, aiming to assess students at different levels in Bloom's taxonomy. Each question is associated with different types of feedback that range from verifying student's answers to giving the student more details related to the answer. The feedback auto-generated in OntoPeFeGe is personalized using a rule-based algorithm which takes into account the student characteristics and the assessment question characteristics. The personalized feedback in OntoPeFeGe was quantitatively evaluated on 88 undergraduate students. The results revealed that the personalized feedback significantly improved the performance of students with low background knowledge. In addition, the feedback was evaluated qualitatively using questionnaires provided to teachers and students. The results showed that teachers and students were satisfied with the feedback.