CNN Learning Research Articles

Forecasting the efficiency of agile projects using machine learning

The aim of the research is to address current challenges in the development of informa-tion systems and the integration of artificial intelligence technologies to optimize Agile meth-odologies. The study focuses on modeling and forecasting the efficiency of Agile processes, taking into account changing requirements over time and random deviations. Forecasting is understood as the process of estimating the future efficiency of Agile processes based on the analysis of time series and machine learning models. Efficiency forecasting is considered as time series forecasting, where parameters adapt depending on the system and data, allowing the model to generalize and take into account various factors affecting project outcomes. The study examines the selection of the optimal model for forecasting Agile process effi-ciency by comparing machine learning methods such as Support Vector Machines (SVM), Random Forest, and neural networks (CNN, LSTM). Comparative analysis showed that neu-ral networks (LSTM, Bidirectional LSTM, CNN) are more effective for predicting project suc-cess, demonstrating high accuracy and lower errors. The practical value of the research lies in identifying LSTM and other neural network architectures as effective tools for predicting project success. This can serve as a guideline for implementing effective management systems in real-world conditions. For example, the application of LSTM in large IT companies for forecasting the success of Agile projects has significantly improved planning accuracy and reduced risks. The research confirms that the integration of machine learning models, particularly LSTM and CNN, significantly improves the accuracy of forecasting and managing Agile pro-jects. The application of these technologies can greatly enhance project management effi-ciency in information systems.

AI-Driven Aircraft Defence: Developing Deep Learning CNN Architectures for Autonomous Systems

AI-Driven Aircraft Defence: Developing Deep Learning CNN Architectures for Autonomous Systems

Metaheuristic-Driven Optimization for Efficient Resource Allocation in Cloud Environments

Intrusion Detection Systems (IDS) play a pivotal role in safeguarding networks against evolving cyber threats. This research focuses on enhancing the performance of IDS using deep learning models, specifically XAI, LSTM, CNN, and GRU, evaluated on the NSL-KDD dataset. The dataset addresses limitations of earlier benchmarks by eliminating redundancies and balancing classes. A robust preprocessing pipeline, including normalization, one-hot encoding, and feature selection, was employed to optimize model inputs. Performance metrics such as Precision, Recall, F1-Score, and Accuracy were used to evaluate models across five attack categories: DoS, Probe, R2L, U2R, and Normal. Results indicate that XAI consistently outperformed other models, achieving the highest accuracy (91.2%) and Precision (91.5%) post-BAT optimization. Comparative analyses of confusion matrices and protocol distributions revealed the dominance of DoS attacks and highlighted specific model challenges with R2L and U2R classes. This study demonstrates the effectiveness of optimized deep learning models in detecting complex attacks, paving the way for robust and adaptive IDS solutions.

Advanced Brain Tumor Classification in MR Images Using Transfer Learning and Pre-Trained Deep CNN Models.

Brain tumor classification is a crucial task in medical diagnostics, as early and accurate detection can significantly improve patient outcomes. This study investigates the effectiveness of pre-trained deep learning models in classifying brain MRI images into four categories: Glioma, Meningioma, Pituitary, and No Tumor, aiming to enhance the diagnostic process through automation. A publicly available Brain Tumor MRI dataset containing 7023 images was used in this research. The study employs state-of-the-art pre-trained models, including Xception, MobileNetV2, InceptionV3, ResNet50, VGG16, and DenseNet121, which are fine-tuned using transfer learning, in combination with advanced preprocessing and data augmentation techniques. Transfer learning was applied to fine-tune the models and optimize classification accuracy while minimizing computational requirements, ensuring efficiency in real-world applications. Among the tested models, Xception emerged as the top performer, achieving a weighted accuracy of 98.73% and a weighted F1 score of 95.29%, demonstrating exceptional generalization capabilities. These models proved particularly effective in addressing class imbalances and delivering consistent performance across various evaluation metrics, thus demonstrating their suitability for clinical adoption. However, challenges persist in improving recall for the Glioma and Meningioma categories, and the black-box nature of deep learning models requires further attention to enhance interpretability and trust in medical settings. The findings underscore the transformative potential of deep learning in medical imaging, offering a pathway toward more reliable, scalable, and efficient diagnostic tools. Future research will focus on expanding dataset diversity, improving model explainability, and validating model performance in real-world clinical settings to support the widespread adoption of AI-driven systems in healthcare and ensure their integration into clinical workflows.

Advancements and Challenges in Character Recognition: A Comparative Analysis of CNN and Deep Learning Approaches

Advancements and Challenges in Character Recognition: A Comparative Analysis of CNN and Deep Learning Approaches

Performance Analysis of MobileNetV3-based Convolutional Neural Network for Facial Skin Disorder Classification

Accurately identifying facial skin types is essential for recommending the right skincare treatments and products. Misidentifying skin types can lead to negative consequences, such as irritation or worsening of skin conditions. This study investigated methods for classifying facial skin types into five categories: oily, acne-prone, dry, normal, and combination. A dataset of 1725 augmented facial images was used. Data augmentation techniques likely increased the dataset's diversity, which helps improve the model's generalization ability. The data underwent preprocessing, including rescaling, before being applied to two deep learning models, CNN and MobileNetV3. The models were evaluated based on accuracy and execution time to determine the most effective approach for classifying facial skin types. The CNN model achieved an accuracy of 64%, demonstrating its potential for image classification tasks. However, the MobileNetV3 model significantly outperformed CNN with an accuracy of 84%. This superior performance is attributed to MobileNetV3's advanced architecture, which is optimized for efficient feature extraction, and particularly relevant for capturing the subtle variations in facial skin types. Therefore, MobileNetV3 emerged as the more effective method for classifying facial skin types with higher accuracy.

Dynamic event-triggered attitude synchronization of multi-spacecraft formation via a learning Chebyshev neural network control approach

PurposeThis paper aims to investigate the attitude synchronization issue of multi-spacecraft formation flying systems under the limited communication resources.Design/methodology/approachThe authors propose a distributed learning Chebyshev neural network controller (LCNNC) combining a dynamic event-triggered (DET) mechanism and a learning CNN model to achieve accurate multi-spacecraft attitude synchronization under communication constraints.FindingsThe proposed method can significantly reduce the internal communication frequency and improve the attitude synchronization accuracy.Practical implicationsThis method requires the low communication resources, has a high control accuracy and is thus suitable for engineering applications.Originality/valueA novel DET mechanism-based LCNNC is proposed to achieve the accurate multi-spacecraft attitude synchronization under communication constraints.

Efficient citrus fruit image classification via a hybrid hierarchical CNN and transfer learning framework

Efficient citrus fruit image classification via a hybrid hierarchical CNN and transfer learning framework

Exploring the Promoter Generation and Prediction of Halomonas spp. Based on GAN and Multi-Model Fusion Methods.

Promoters, as core elements in the regulation of gene expression, play a pivotal role in genetic engineering and synthetic biology. The accurate prediction and optimization of promoter strength are essential for advancing these fields. Here, we present the first promoter strength database tailored to Halomonas, an extremophilic microorganism, and propose a novel promoter design and prediction method based on generative adversarial networks (GANs) and multi-model fusion. The GAN model effectively learns the key features of Halomonas promoter sequences, such as the GC content and Moran's coefficients, to generate biologically plausible promoter sequences. To enhance prediction accuracy, we developed a multi-model fusion framework integrating deep learning and machine learning approaches. Deep learning models, incorporating BiLSTM and CNN architectures, capture k-mer and PSSM features, whereas machine learning models utilize engineered string and non-string features to construct comprehensive feature matrices for the multidimensional analysis and prediction of promoter strength. Using the proposed framework, newly generated promoters via mutation were predicted, and their functional validity was experimentally confirmed. The integration of multiple models significantly reduced the experimental validation space through an intersection-based strategy, achieving a notable improvement in top quantile prediction accuracy, particularly within the top five quantiles. The robustness and applicability of this model were further validated on diverse datasets, including test sets and out-of-sample promoters. This study not only introduces an innovative approach for promoter design and prediction in Halomonas but also lays a foundation for advancing industrial biotechnology. Additionally, the proposed strategy of GAN-based generation coupled with multi-model prediction demonstrates versatility, offering a valuable reference for promoter design and strength prediction in other extremophiles. Our findings highlight the promising synergy between artificial intelligence and synthetic biology, underscoring their profound academic and practical implications.

FedHFP: A Federated Deep Learning Framework for Heart Failure Prediction

Heart failure emerges as one of the leading global causes of mortality, presenting a significant challenge in healthcare that necessitates early detection and intervention. Recent advances in Federated Learning (FL) have shown substantial potential in enhancing predictive analytics for heart failure, potentially enabling more timely interventions that could reduce the risk of heart failure. However, despite its promising potential, FL in healthcare is often complicated by technical challenges, such as model training across decentralized networks with limited infrastructure. Additionally, there is a gap in the literature, as no notable study has provided a comparative analysis of FL models under diverse settings. This paper proposes the FedHFP, a Federated deep learning framework for heart failure prediction, emphasizing its potential impact and utility in remote regions with limited healthcare access. Further, we present a comprehensive comparison of various deep learning models, including ANN, CNN, RNN, LSTM, and GRU, aimed at leveraging FL to optimize heart failure prediction. Through comparative analysis across diverse network configurations, the findings indicate that ANN may outperform other models, achieving an accuracy of ∼93.75%. Furthermore, our study emphasizes, at a conceptual level, the application of these findings to remote areas with limited healthcare infrastructure. The results underscore the efficacy of FL in healthcare predictive modeling and highlight its potential to mitigate communication overhead and preserve data privacy, particularly benefiting regions with limited healthcare infrastructure.

An Enhanced Feature Learning CNN Using Auto-encoder for Partially Occluded Target Recognition in SAR Imagery

Deep learning algorithms have found extensive utility in Automatic Target Recognition (ATR) using Synthetic Aperture Radar (SAR). Despite this, conventional deep learning-based ATR models often lack in their performance when subjected to Extended Operating Conditions (EOCs). Significant variations in depression, noise interference, limited training data, and partial occlusion are primary factors in EOCs. In this paper, we proposed a deep neural network-based technique to recognize SAR targets for Standard Operating Conditions (SOC) and Extended Operating Conditions (EOC). The proposed approach comprises an encoder and decoder network. The encoder network integrates multiple dilated convolutions to address noise-induced recognition challenges for extracting multi-scale features. Furthermore, a feature refinement module is integrated within the multi-scale channels to enhance the extraction of discriminative and resilient features. Feature refinement selectively emphasizes informative features while suppressing irrelevant ones. Additionally, a salient attention block and a spatial feature learning module are introduced for feature selection. Selected features are then utilized in the top-level network layer of the encoder network to preserve spatial relationships among diverse features. This spatial feature learning module aids in recognition performance, particularly when training data is scarce. The decoder network consists of stacked transposed convolution layers, facilitating the encoder network's acquisition of discriminative image features. Experimental evaluations were conducted on the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset to validate the efficacy of the proposed methodology.

Simultaneous monitoring of two comprehensive quality evaluation indexes of frozen-thawed beef meatballs using hyperspectral imaging and multi-task convolutional neural network

The quality of beef meatballs during repeated freeze-thaw (F-T) cycles was assessed by multiple indicators. This study introduced a novel quality evaluation method using hyperspectral imaging (HSI) and multi-task learning. Seventeen quality indicators were analyzed to assess the impact of F-T cycles. Subsequently, a comprehensive quality index (CQI) and a comprehensive weight index (CWI) were constructed from 11 key indicators via factor analysis. By integrating HSI data from 150 samples with multi-task convolutional neural network (MT-CNN), the feasibility of simultaneous monitoring of CQI and CWI of the beef meatballs was explored. The results demonstrated that MT-CNN achieved superior predictions for CQI (RMSEp = 1.24, R2 = 0.94) and CWI (RMSEp = 20.436, R2 = 0.94) compared to traditional machine learning and single-task CNN approaches. Furthermore, the deterioration trends of beef meatballs during multiple F-T cycles were effectively visualized. Thus, the integration of HSI and MT-CNN enabled efficient prediction of comprehensive evaluation indexes for beef meatballs, contributing to their quality control.

A Multi-Scale CNN for Transfer Learning in sEMG-Based Hand Gesture Recognition for Prosthetic Devices.

Advancements in neural network approaches have enhanced the effectiveness of surface Electromyography (sEMG)-based hand gesture recognition when measuring muscle activity. However, current deep learning architectures struggle to achieve good generalization and robustness, often demanding significant computational resources. The goal of this paper was to develop a robust model that can quickly adapt to new users using Transfer Learning. We propose a Multi-Scale Convolutional Neural Network (MSCNN), pre-trained with various strategies to improve inter-subject generalization. These strategies include domain adaptation with a gradient-reversal layer and self-supervision using triplet margin loss. We evaluated these approaches on several benchmark datasets, specifically the NinaPro databases. This study also compared two different Transfer Learning frameworks designed for user-dependent fine-tuning. The second Transfer Learning framework achieved a 97% F1 Score across 14 classes with an average of 1.40 epochs, suggesting potential for on-site model retraining in cases of performance degradation over time. The findings highlight the effectiveness of Transfer Learning in creating adaptive, user-specific models for sEMG-based prosthetic hands. Moreover, the study examined the impacts of rectification and window length, with a focus on real-time accessible normalizing techniques, suggesting significant improvements in usability and performance.

Exploring the Influence of Tropical Cyclones on Regional Air Quality Using Multimodal Deep Learning Techniques.

Tropical cyclones (TC) are dynamic atmospheric phenomena featuring extreme low-pressure systems and powerful winds, known for their devastating impacts on weather and the environment. The main purpose of this paper is to consider the subtle involvement of TCs in the air quality index (AQI), focusing on aspects related to the air quality before, during and after cyclones. This research employs multimodal methods, which include meteorological data and different satellite observations. Deep learning approaches, i.e., ConvLSTM, CNN and Real-ESRGAN models, are combined with a regression model to analyze the temporal variability in the air quality associated with tropical cyclones. Deep learning models are deployed to uncover complex patterns and non-linear interdependencies between cyclones' features and the AQI to give predictive insights into the air quality fluctuations throughout the different stages of tropical cyclones. Furthermore, this study explores the aftermaths of TCs in terms of the air quality with respect to post-cyclone recovery. The findings offer an enhanced view of the role of TCs in the regional or global air quality, which will be useful for policymakers, meteorologists and environmental researchers. Utilizing a CNN for tropical cyclone (TC) classification and the extra trees regressor (ETR) for AQI prediction results in accuracy of 92.02% for the CNN and an R2 of 83.33% for the ETR. Hence, this work adds to our knowledge and enlightens us on the complex interactions between TCs and the air quality, highlighting wider public health concerns regarding climate adaptation and urban renewal.

BEES: Bio-Inspired Enhanced Emotional Segmentation System for Depression Detection using Deep Learning CNN with Bee Colony Optimization

BEES: Bio-Inspired Enhanced Emotional Segmentation System for Depression Detection using Deep Learning CNN with Bee Colony Optimization

Revolutionizing emergency vehicle response: a smart traffic management approach with adaptive CNN and hybrid deep learning

Revolutionizing emergency vehicle response: a smart traffic management approach with adaptive CNN and hybrid deep learning

Offline English Handwritten Character Recognition using Sequential Convolutional Neural Network

Handwritten Character recognition falls under the domain of image classification that has been under research for years. The idea is to make the machine recognize handwritten human characters. The language focused in this research paper is English while using offline handwritten character recognition for identifying English characters. There are many publically available datasets from which EMNIST is the most challenging one. The main idea of this research paper is to propose a deep learning CNN method to help recognize English characters. This research paper proposes a deep learning convolutional neural network that is tested and compared with renowned pre-trained models using transfer learning. These parametric settings address multiple issues and are finalized after experimentation. The same hyper-parametric settings were used for all the models under test and E-Character with the same data augmentation settings. The proposed model named the E-Character recognizer was able to produce 87.31% accuracy. It was better than most of the tested pre-trained models and other proposed methods by other researchers. This research paper further highlighted some of the problems like misclassification due to the similar structure of characters.

Machine learning based Parkinson’s disease detection and progression evaluation using gray and white matter segmentation from 3D MRI

Parkinson’s disease (PD) is a movement-related neurological condition caused by the death of brain nerve cells that produce dopamine. T1 MR images were obtained from the Parkinson’s Progression Markers Initiative (PPMI) database. Data was collected at baseline, and at 48 months duration. SPM12 software was used to segment gray matter (GM) and white matter (WM) from the MR images. For the classification of PD, CNN and machine learning (ML) classifiers were used to train on the segmented GM and WM. The findings demonstrated that utilizing the segmented GM and WM obtained from MR images at 48 months had a better predictive ability than the data obtainment at the baseline. CNNs did not perform as well as the conventional ML algorithms, especially for the baseline data. This result is most likely due to the smaller dataset available for training the CNNs, as CNNs normally require more data for best performance. Overall, machine learning algorithms were able to distinguish between people with Parkinson’s disease and healthy controls by analyzing GM and WM in brain scans. The classifiers trained at different stages demonstrated varying degrees of accuracy, with the predictive accuracy for the 48-month dataset surpassing that of the baseline data by a significant margin. The accuracy rate for GM was 65.78% at baseline and improved significantly to 92.59% at 48 months. Similarly, the accuracy rate for WM was 60.52% at baseline and improved to 88.89% at 48 months.

DOA Estimation: LSTM and CNN Learning Algorithms

DOA Estimation: LSTM and CNN Learning Algorithms

Malaysian Noodle Images Classification System Using CNN and Transfer Learning

Malaysian Noodle Images Classification System Using CNN and Transfer Learning