Neural Computing and Applications最新文献

Skin cancer poses a significant health hazard, necessitating the utilization of advanced diagnostic methodologies to facilitate timely detection, owing to its escalating prevalence in recent years. This paper proposes a novel approach to tackle the issue by introducing a method for detecting skin cancer that uses MixNets to enhance diagnosis and leverages mobile network-based transfer learning. Skin cancer has diverse forms, each distinguishable by its structural attributes, morphological characteristics, texture, and coloration. The pressing demand for accurate and efficient diagnostic instruments has spurred the investigation of novel techniques. The present study utilizes the ISIC dataset, comprising a validation set of 660 images and a training set of 2637 images. Moreover, the research employs a combination of MixNets and mobile network-based transfer learning as its chosen approach. Transfer learning is a technique that leverages preexisting models to enhance the diagnostic capabilities of the proposed system. Integrating MobileNet and MixNets allows for utilizing their respective functionalities, resulting in a dual-model methodology that enhances the comprehensiveness of skin cancer diagnosis. The results demonstrate impressive performance metrics, with MobileNet and MixNets models, and the proposed approach achieves an outstanding accuracy rate of 99.58%. The above findings underscore the efficacy of the dual-model method in effectively discerning between benign and malignant skin lesions. Moreover, the present study aims to examine the potential integration of emerging technologies to enhance the accuracy and practicality of diagnostics within real-world healthcare settings.

Ethereum has emerged as a major platform for decentralized apps and smart contracts with the heightened interest in cryptocurrencies in recent years. Investors and market participants in the cryptocurrency space will find it increasingly important to use reliable price prediction models as Ethereum's popularity grows. To better estimate Ethereum prices around the world, we propose "EtherVoyant," a novel hybrid forecasting model that combines the advantages of ARIMA and SARIMA methods. To improve its forecasting abilities, EtherVoyant uses Ethereum price history to train ARIMA and SARIMA components independently before fusing their predictions. With the help of feature engineering and data preparation, we further improve the model so that it can deal with real-world difficulties like missing values and seasonality in the data. We also investigate hyperparameter optimization for the model's best possible performance. We compare EtherVoyant's forecasts against those of the more conventional ARIMA and SARIMA models to determine its efficacy. By providing more precise and trustworthy price forecasts, our trial results suggest that EtherVoyant is superior to the individual models. The importance of this study resides in the fact that it will lead to the creation of a sophisticated time series forecasting model that will be useful to cryptocurrency investors, traders, and decision-makers. We hope that by making EtherVoyant available on a worldwide scale, we will help advance the field of cryptocurrency analytics and encourage wider adoption of blockchain-based assets.

This paper presents a novel optimization algorithm called hyper-heuristic fitness-distance balance success-history-based adaptive differential evolution (HH-FDB-SHADE). The hyper-heuristic algorithms have two main structures: a hyper-selection framework and a low-level heuristic (LLH) pool. In the proposed algorithm, the FDB method is preferred as a high-level selection framework to evaluate the LLH pool algorithms. In addition, a total of 10 different strategies is derived from five mutation operators and two crossover methods for using them as the LLH pool. Balancing the exploration and exploitation capability of FDB is the main reason for being the selection framework of the proposed algorithm. The success of the HH-FDB-SHADE algorithm was tested on CEC-17 and CEC-20 benchmark test suits for different dimensional search spaces, and the obtained solutions from the HH-FDB-SHADE were compared to 10 different LLH pool algorithms. In addition, the HH-FDB-SHADE algorithm has been applied to optimize the control parameters of PID, PIDF, FOPID, and PIDD² in the optimal automatic voltage regulator (AVR) design problem to reveal the improved algorithm's performance more clearly and prove its success in solving engineering problems. The results obtained from the AVR system are compared with five other effective meta-heuristic search algorithms such as the fitness-distance balance Lévy Flight distribution, differential evolution, Harris–Hawks optimization, Barnacles mating optimizer, and Moth–Flame optimization algorithms in the literature. The results of the statistical analyses indicate that HH-FDB-SHADE is the best-ranked algorithm for solving CEC-17 and CEC-20 benchmark problems and gives better results compared to the LLH pool algorithms. Besides, the proposed algorithm is more effective and robust than five other meta-heuristic algorithms in solving optimal AVR design problems.

Early detection of microsatellite instability (MSI) and microsatellite stability (MSS) is crucial in the fight against gastrointestinal (GI) cancer. MSI is a sign of genetic instability often associated with DNA repair mechanism deficiencies, which can cause (GI) cancers. On the other hand, MSS signifies genomic stability in microsatellite regions. Differentiating between these two states is pivotal in clinical decision-making as it provides prognostic and predictive information and treatment strategies. Rapid identification of MSI and MSS enables oncologists to tailor therapies more accurately, potentially saving patients from unnecessary treatments and guiding them toward regimens with the highest likelihood of success. Detecting these microsatellite status markers at an initial stage can improve patient outcomes and quality of life in GI cancer management. Our research paper introduces a cutting-edge method for detecting early GI cancer using deep learning (DL). Our goal is to identify the optimal model for GI cancer detection that surpasses previous works. Our proposed model comprises four stages: data acquisition, image processing, feature extraction, and classification. We use histopathology images from the Cancer Genome Atlas (TCGA) and Kaggle website with some modifications for data acquisition. In the image processing stage, we apply various operations such as color transformation, resizing, normalization, and labeling to prepare the input image for enrollment in our DL models. We present five different DL models, including convolutional neural networks (CNNs), a hybrid of CNNs-simple RNN (recurrent neural network), a hybrid of CNNs with long short-term memory (LSTM) (CNNs-LSTM), a hybrid of CNNs with gated recurrent unit (GRU) (CNNs-GRU), and a hybrid of CNNs-SimpleRNN-LSTM-GRU. Our empirical results demonstrate that CNNs-SimpleRNN-LSTM-GRU outperforms other models in accuracy, specificity, recall, precision, AUC, and F1, achieving an accuracy of 99.90%. Our proposed methodology offers significant improvements in GI cancer detection compared to recent techniques, highlighting the potential of DL-based approaches for histopathology data. We expect our findings to inspire future research in DL-based GI cancer detection.

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.

In areas where conventional energy sources are unavailable, alternative energy technologies play a crucial role in generating electricity. These technologies offer various benefits, such as reliable energy supply, environmental sustainability, and employment opportunities in rural regions. This study focuses on the development of a novel optimization algorithm called the modified firebug swarm algorithm (mFSO). Its objective is to determine the optimal size of an integrated renewable power system for supplying electricity to a specific remote site in Dehiba town, located in the eastern province of Tataouine, Tunisia. The proposed configuration for the standalone hybrid system involves PV/biomass/battery, and three objective functions are considered: minimizing the total energy cost (COE), reducing the loss of power supply probability (LPSP), and managing excess energy (EXC). The effectiveness of the modified algorithm is evaluated using various tests, including the Wilcoxon test, boxplot analysis, and the ten benchmark functions of the CEC2020 benchmark. Comparative analysis between the mFSO and widely used algorithms like the original Firebug Swarm Optimization (FSO), Slime Mold Algorithm (SMA), and Seagull Optimization Algorithm (SOA) demonstrates that the proposed mFSO technique is efficient and effective in solving the design problem, surpassing other optimization algorithms.

The malfunctioning of cardiac autonomic control in epileptic patients develops ventricular tachyarrhythmia and causes sudden unexpected death in epilepsy patients (SUDEP). Various clinical studies investigated the effect of epilepsy on cardiac autonomic control by performing heart rate variability (HRV) analysis; however, results are unclear regarding whether sympathetic, parasympathetic, or both branches of the autonomic nervous system (ANS) are affected in epilepsy and also the impact of anticonvulsant treatment on the ANS. This study follows the systematic protocols to investigate epilepsy and its anticonvulsant treatment on cardiac autonomic control by using linear and nonlinear HRV analysis measures. The electronic databases of PubMed, Embase, and Cochrane Library were used for the collection of studies. Initially, 1475 articles were identified whereas after 2-staged exclusion criteria, 33 studies were selected for execution of the review process and meta-analysis. For meta-analysis, four comparisons were performed (epilepsy patients): (1) controls (healthy subject with no history of epilepsy) versus untreated patients; (2) treated (patients under treatment that have a seizure) versus untreated patients; (3) controls versus treated patients; and (4) refractory versus well-controlled (epilepsy patients that were seizure-free for last 1 year). For treated and untreated patients, there was no significant difference whereas well-controlled patients presented higher values as compared to refractory patients. Meta-analysis was performed for the time-domain, frequency-domain, and nonlinear parameters. Untreated patients in comparison with controls presented significantly lower HF (high-frequency) and LF (low-frequency) values. These LF (g = − 0.9; 95% CI − 1.48 to − 0.37) and HF (g = − 0.69; 95% confidence interval (CI) − 1.24 to − 0.16) values were affirming suppressed both, vagal and sympathetic activity, respectively. Additionally, LF and HF value was increased in most of the studies indicating suppressed vagal tone, while for some studies, their value decreased to indicate suppressed sympathetic activity. No significant difference was observed for the remaining comparisons. Results affirmed the hypothesis that suppressed sympathetic activity affects sympathovagal balance and leads to SUDEP, as the LF value was significantly lower for patients as compared to healthy subjects. The overall effect size and statistical results for LF and HF were significant, showing the research and clinical significance of our study.

Time series forecasting models are essential decision support tools in real-world domains. Stock market is a remarkably complex domain, due to its quickly evolving temporal nature, as well as the multiple factors having an impact on stock prices. To date, a number of machine learning-based approaches have been proposed in the literature to tackle stock trend prediction. However, they typically tend to analyze a single data source or modality, or consider multiple modalities in isolation and rely on simple combination strategies, with a potential reduction in their modeling power. In this paper, we propose a multimodal deep fusion model to predict stock trends, leveraging daily stock prices, technical indicators, and sentiment in daily news headlines published by media outlets. The proposed architecture leverages a BERT-based model branch fine-tuned on financial news and a long short-term memory (LSTM) branch that captures relevant temporal patterns in multivariate data, including stock prices and technical indicators. Our experiments on 12 different stock datasets with prices and news headlines demonstrate that our proposed model is more effective than popular baseline approaches, both in terms of accuracy and trading performance in a portfolio analysis simulation, highlighting the positive impact of multimodal deep learning for stock trend prediction.

Unmanned Aerial Vehicles (UAVs) have been gaining popularity in various commercial, civilian, and military applications due to their efficiency and cost-effectiveness. However, the increasing demand for UAVs makes them vulnerable to various cyberattacks/intrusions that could have devastating consequences at an individual, organizational, and national level. To mitigate this, prompt detection of such threats is crucial in order to prevent potential damage and ensure safe and secure operations. In this work, we provide an overview of UAV systems’ architecture, security, and privacy requirements. We then analyze potential threats to UAVs, providing an evaluation of countermeasures for UAV-based attacks. We also present a comprehensive and timely exploration of state-of-the-art UAV Intrusion Detection Systems (IDSs), specifically focusing on Machine Learning (ML)-based approaches. We look at the increasing importance of using ML for detecting intrusions in UAVs, which have gained significant attention from both academia and industry. This study also takes a step forward by pointing out and classifying contemporary IDSs based on their detection methods, feature selection techniques, evaluation datasets, and performance metrics. By evaluating existing research, we aim to provide more insight into the issues and limitations of current UAV IDSs. Additionally, we identify research gaps and challenges while suggesting potential future research directions in this domain.

Radial basis function (RBF) neural network training presents a challenging optimization task, necessitating the utilization of advanced algorithms that can fully train the network so as to produce accurate and computationally efficient models. To achieve this goal, this work introduces a new framework where the original RBF training problem is divided into two simpler subproblems; the linear parameters, namely the network weights, are projected out of the problem using variable projection (VP), thus leaving a reduced functional, which depends only on nonlinear parameters, i.e., the RBF centers. The centers are updated using the Levenberg–Marquardt (LM) algorithm, while the optimal values of the synaptic weights are calculated in each iteration of the LM algorithm using linear regression. The proposed VP-LM scheme is coupled with the fuzzy means (FM) algorithm, which helps to select the number of RBF centers and enhances the overall search procedure, thus resulting to a framework that produces parsimonious models with enhanced accuracy in shorter training times. The proposed training scheme is evaluated on 12 both real-world and synthetic benchmark datasets and tested against various RBF training algorithms, as well as different neural network architectures. The experimental results underscore the effectiveness of the VP-FM algorithm in producing neural network models that outperform those generated by alternative methods in many aspects; to be more specific, the proposed approach achieves very competitive model accuracy, while resulting to smaller network sizes and thus lower complexity, which leads to shorter training times.