Search Results (411)

This paper proposes a multi-strategy-enhanced Octopus Optimization Algorithm (OOA) for hyperparameter optimization in hyperspectral image classification. Hyperspectral images pose significant challenges due to their numerous spectral bands, high dimensionality, and complex spectral differences between classes, which complicate classification modeling. The classification performance of support vector machine (SVM) classifiers is also highly dependent on parameter settings. The original OOA is extended by incorporating an initialization strategy based on elite backpropagation, a multi-stage nonlinear adaptive parameter control mechanism, an elite-guided differential mutation strategy, a Lévy flight restart mechanism with stagnation monitoring, and a stable boundary handling strategy. These enhancements constitute the IOOA-SVM parameter optimization framework. The proposed method is evaluated against OOA, Particle Swarm Optimization (PSO), Sand Cat Swarm Optimization (SCSO), Salp Swarm Algorithm (SSA), Grey Wolf Optimizer (GWO), Arithmetic Optimization Algorithm (AOA), Differential Evolution (DE) and Linear Population Size Reduction Success-History Based Adaptive Differential Evolution (L-SHADE) on the CEC2017 test set, achieving superior results on most of the 29 test functions, IOOA achieved the best results on average for 27 of the 29 test functions, outperforming the original OOA on all 29 test functions and demonstrating superior performance on most stability metrics. Different improvement strategies yield varying degrees of performance gains for the algorithm; among them, the elite-guided differential mutation strategy produces the most significant performance improvement. The synergy and complementarity among multiple strategies play a major role in enhancing the performance of the Improved Octopus Optimization Algorithm. Experimental results show that the SVM classifier optimized using the improved OOA achieves a classification accuracy of 97.3731%, representing a 0.2278 percentage point improvement over the original algorithm and demonstrating strong overall optimization performance. Full article

Hybrid quantum–classical machine learning offers a promising direction for advancing automated quality control in industrial settings. In this study, we investigate two hybrid quantum–classical approaches for classifying defects in aluminum TIG welding images and benchmarking their performance against a conventional deep learning model. A convolutional neural network is used to extract compact and informative feature vectors from weld images, effectively reducing the higher-dimensional pixel space to a lower-dimensional feature space. Our first quantum approach encodes these features into quantum states using a parameterized quantum feature map composed of rotation and entangling gates. We compute a quantum kernel matrix from the inner products of these states, defining a linear system in a higher-dimensional Hilbert space corresponding to the support vector machine (SVM) optimization problem and solving it using a Variational Quantum Linear Solver (VQLS). We also examine the effect of the quantum kernel condition number on classification performance. In our second method, we apply angle encoding to the extracted features in a variational quantum circuit and use a classical optimizer for model training. Both quantum models are tested on binary and multiclass classification tasks, and the performance is compared with the classical CNN model. Our results show that while the CNN model demonstrates robust performance, hybrid quantum–classical models perform competitively. This highlights the potential of hybrid quantum–classical approaches for near-term real-world applications in industrial defect detection and quality assurance. Full article

Background: Anaplastic Lymphoma Kinase (ALK) is an oncogenic receptor tyrosine kinase implicated in several cancers. Despite the clinical success of ALK inhibitors, acquired resistance continues to drive the search for novel chemotypes. We developed a multiclass machine learning framework to classify ALK inhibitory activity using a curated ChEMBL dataset. Methods: Models were built using 2D molecular descriptors together with MACCS and ECFP4 fingerprints. Three widely used algorithms, Support Vector Machine (SVM), Random Forest (RF), and XGBoost, were applied for model development. Results: RF and XGBoost models demonstrated the best performance, achieving accuracies of ~0.75–0.79 with consistently high ROC–AUC values, particularly for fingerprint-based features. Bemis–Murcko scaffold analysis identified enriched chemotypes and underexplored scaffolds for further prioritization. The validated models were subsequently used to screen the Maybridge library, and compounds predicted to possess potential ALK inhibitory activity were prioritized for further computational evaluation. Applicability-domain filtering confirmed that the selected compounds occupied the predicted ALK inhibitor chemical space across multiple activity classes. The shortlisted compounds were subsequently evaluated by molecular docking to characterize their binding modes and interactions. Three candidate hits (SCR00078, SCR00073, and AW01085) were selected for further evaluation using 500 ns molecular dynamics simulations alongside the reference inhibitor Brigatinib. Simulation analyses revealed stable protein–ligand complexes and reduced conformational fluctuations relative to apo ALK, while MM/PBSA calculations identified SCR00078 and AW01085 as the most favorable binders. Conclusions: This integrated ML-to-simulation workflow prioritizes structurally novel candidate hits with predicted ALK inhibitory activity and provides an effective strategy for scaffold discovery and hit prioritization. Full article

When marine chillers operate under complex marine conditions, they suffer from severe cross-equipment feature distribution shifts, scarce labeled fault samples in the target domain, industrial vibration noise mixed in sensor signals, and difficulties in accurately identifying subtle faults with varying severity levels. To tackle these issues, this paper improves upon the domain difference perception network (DDPN) and proposes a dual-hybrid attention feature discriminant domain-Adversarial network (DAFDAN) to realize intelligent fault diagnosis across different equipment and working conditions under few-shot scenarios. The proposed method constructs a dual-branch feature encoder consisting of a source domain compressor and a target domain extender to accommodate the distinct sensor dimensions of two heterogeneous chiller types. A hybrid attention module is formed by integrating squeeze-and-excitation efficient channel attention (SE-ECA, a module for screening channel-wise features) and spatial attention, which adaptively amplifies time-series features sensitive to faults and suppresses irrelevant noise. Residual connections (shortcut paths in deep neural networks to mitigate the vanishing gradient problem during deep-layer training) are introduced to optimize feature transmission. A dual-layer domain alignment framework is built with gradient reversal layers and maximum mean discrepancy (MMD). Combined with adversarial training (a training paradigm that learns domain-agnostic features through a game between a feature extractor and a domain discriminator), the framework achieves joint optimization of implicit feature confusion and explicit distance constraints. Meanwhile, a five-stage progressive training strategy is designed, which activates multiple loss functions, including weighted cross-entropy, mean square error (MSE), binary cross-entropy (BCE), and Kullback–Leibler (KL) divergence stage by stage. Class weighting and early stopping strategies are adopted to alleviate sample imbalance and model overfitting. In this paper, the public ASHRAE RP-1043 centrifugal chiller dataset is used as the source domain, and time-series measurement data collected from a self-developed laboratory marine screw chiller serves as the target domain. Verification experiments are carried out covering one normal steady-state operating condition and 15 gradient faults falling into five major categories with different severity degrees. Results from ablation experiments (controlled-variable comparative experiments that quantify the independent contribution of each component by comparing model performance with or without a specific module/loss), multi-algorithm comparisons, and confusion matrix visualization demonstrate that the cross-domain fault diagnosis accuracy of the proposed DAFDAN approaches is 100%, outperforming mainstream transfer learning algorithms such as support vector machine (SVM), deep neural network (DNN), MMD, correlation alignment (CORAL), and domain-adversarial neural network (DANN). Multiple ablation experiments verify that the three core components—hybrid attention, adversarial training, and semi-supervised learning—jointly boost the model’s diagnosis accuracy and operational stability. The loss curves of the complete five-stage training process converge smoothly. The confusion matrix reveals zero misjudgments and zero false alarms across all 16 refined operating states, enabling precise identification of subtle incipient faults of all severity levels. This study proves that DAFDAN can effectively address the pain points of few-shot cross-equipment fault diagnosis for marine chillers and provides a reliable algorithmic reference for the intelligent operation and maintenance of ship refrigeration equipment. Full article

Background/Objectives: Obstructivesleep apnea (OSA) is one of the most widespread forms of sleep disease, affecting over 936 million adults globally. The health consequences of obstructive sleep apnea (OSA) are well documented; however, it remains largely underdiagnosed because the current gold-standard diagnostic method, polysomnography (PSG), is often costly, time-consuming, and unavailable in many healthcare settings. To address these challenges, this study presents a novel explainable deep learning (DL) framework for automated multi-class OSA severity classification using single-lead electrocardiogram (ECG) signals. Methods: The proposed framework integrates a hybrid CNN–BiLSTM architecture with explainable artificial intelligence (XAI) techniques to generate clinically meaningful predictions and explanations across four OSA severity classes: Normal, Mild, Moderate, and Severe. The framework was evaluated using the publicly available PhysioNet Apnea-ECG dataset (70 recordings) together with an institutional ECG dataset (150 recordings), resulting in a combined cohort of 220 recordings. Results: The proposed framework achieved an overall classification accuracy of 94.7%, with sensitivity and specificity values of 92.3% and 96.1%, respectively. Furthermore, the proposed model consistently outperformed conventional machine learning algorithms, including Support Vector Machine (SVM), Random Forest, and XGBoost, by 5.5%, 4.2%, and 2.9%, respectively. To enhance transparency and clinical trust, SHAP (SHapley Additive exPlanations) was employed to identify the most influential physiological predictors driving model decisions. Heart rate variability features, particularly RMSSD and pNN50, emerged as the strongest indicators of OSA severity. Moreover, computational efficiency analysis revealed that the model required only 0.23 s to process a 60 s ECG epoch on a standard computing platform, supporting its suitability for real-time deployment. Conclusions: The findings demonstrate that explainable deep learning applied to ECG signals can provide accurate, interpretable, and computationally efficient assessment of OSA severity. The proposed framework may support OSA screening, clinical triage, and early intervention, particularly in resource-constrained healthcare environments. Full article

Potato leaf disease poses a significant threat to global food security, causing substantial crop losses that jeopardise agricultural productivity and farmers’ livelihoods worldwide. Existing automated detection frameworks suffer from several persistent limitations, including over-reliance on controlled benchmark datasets, narrow disease class coverage, exclusive use of spatial feature representations, absence of feature selection, and dependence on single-architecture end-to-end pipelines. To address these limitations, this paper proposes ComPo-Net, a novel lightweight ensemble framework that integrates three efficient CNN architectures—ResNet18, ShuffleNet, and MobileNetV2—for nine-class potato leaf disease detection and classification. Deep features are extracted from three intermediate layers of each network, with the Discrete Wavelet Transform applied for dimensionality reduction and cross-network fusion of the higher-dimensional layer features, capturing spectral–spatial information that purely spatial approaches cannot provide, while the remaining layer features are directly concatenated across networks. One-way Analysis of Variance (ANOVA) feature selection is subsequently applied to retain the most statistically significant features from the combined multi-scale, multi-network representation, and seven machine learning classifiers are systematically evaluated to identify the optimal classification strategy. The framework is assessed on a merged dataset of three publicly available benchmarks spanning both controlled and uncontrolled imaging environments, constituting a nine-class evaluation setting not previously addressed at this scale in the literature. ComPo-Net achieves an accuracy of 96.32%, an F1-score of 93.87%, an MCC of 0.9350, and AUC values exceeding 0.993 across all nine classes with Cubic SVM as the best-performing classifier. When compared against methods evaluated on the seven-class uncontrolled-environment dataset—the closest available task setting to ComPo-Net’s nine-class merged benchmark—ComPo-Net surpasses the best-performing comparable method by a margin of 6.45 percentage points, demonstrating the effectiveness of multi-scale ensemble feature extraction combined with spectral–spatial representation and principled feature selection for robust potato leaf disease detection under diverse real-world conditions. Full article

High-precision mapping of tree species composition is essential for sustainable forest management, biodiversity assessment, and ecosystem monitoring. Airborne hyperspectral imagery provides rich spectral and spatial information that enables detailed species discrimination. However, traditional single-stream convolutional neural networks (CNNs) often fail to fully exploit multi-dimensional features and are susceptible to spectral redundancy. In this study, we propose the MS-CBAM-TSCNet (Multi-Stream Convolutional Block Attention Deep Neural Network), a novel architecture specifically designed for tree species classification using aerial hyperspectral data. The proposed model integrates three parallel processing streams: a 1D spectral branch for capturing reflectance signatures, a 2D spatial branch for modeling contextual patterns, and a 3D spectral–spatial branch enhanced with Convolutional Block Attention Modules (CBAMs) to adaptively recalibrate channel-wise and spatial–spectral features. An adaptive gated fusion mechanism with attention-based weighting is introduced to dynamically combine the complementary representations extracted from the three streams, improving robustness to spectral redundancy and class imbalance. The method was evaluated on an airborne HyMap hyperspectral dataset (125 bands, with 4 m spatial resolution) acquired over a mixed boreal forest in Karlsruhe, Germany, comprising five dominant tree species. Using five-fold cross-validation on an augmented dataset, the MS-CBAM-TSCNet achieved an overall accuracy of 96.8%, a Kappa coefficient of 0.96, and a macro F1-score of 0.966, outperforming conventional 1D, 2D, and 3D CNNs, as well as a hybrid CNN-SVM approach across all evaluation metrics. An ablation study further confirms the complementary contributions of the multi-stream architecture, CBAM attention, and adaptive gated fusion. Pixel-wise classification maps demonstrate improved boundary delineation and reduced misclassification in mixed stands, highlighting the effectiveness of the proposed framework for operational forest inventory and ecological monitoring. Full article

The rapid decline of Arctic sea ice is increasing the feasibility of trans-Arctic shipping along the Northern Sea Route (NSR)—a considerably shorter and potentially lower-emission alternative to the Suez Canal route—for which the Bering Strait is the first gateway chokepoint for vessels departing the Republic of Korea. Continuous optical monitoring of this region is fundamentally limited by cloud cover and polar night during the ice season; in 2025, no usable Sentinel-2 MultiSpectral Instrument (MSI) scene was available for January (mean cloud cover 56.9% in November). This study therefore used all-weather Sentinel-1 dual-polarization (VV, VH) C-band SAR, acquired monthly through Google Earth Engine for January–December 2025, to classify open water, thin ice (<30 cm), and first-year ice (FYI, ≥30 cm) with a Support Vector Machine (SVM); land was masked prior to classification. Three kernels (linear, radial basis function (RBF), and polynomial) were compared after $5\times 5$ Lee speckle filtering, using VV, VH, and the VV/VH ratio (in dB) as input features. The RBF kernel achieved the highest accuracy (overall accuracy 97.0%, $\kappa =0.954$), and the classification was cross-referenced against the U.S. National Snow and Ice Data Center (NSIDC) Multisensor Analyzed Sea Ice Extent (MASIE) mask. The monthly ice maps were then reclassified into World Meteorological Organization (WMO)/POLARIS ice types and converted into a POLARIS Risk Index Outcome (RIO) cost surface using simplified proxy weights consistent with the POLARIS risk ordering rather than the full Risk Index Value table; least-cost paths and a seasonal navigability assessment were derived for three representative ship ice classes—a non-ice-class vessel, an ice-class 1A merchant vessel (≈PC7), and the icebreaker Araon (≈PC5)—under risk-index, water-depth, and coastal-buffer constraints. For a representative refreezing-onset scene, the routing outcome was strongly ice-class dependent: the non-ice-class vessel and the ice-class 1A vessel were blocked (after 42 and 139 km, respectively), whereas only the icebreaker Araon completed the transit along a 124 km least-cost path within a 2 km safety corridor. The framework demonstrates how an interpretable SAR-based sea-ice classification can be coupled with a recognized international risk standard to support risk-informed navigation through the Bering Strait gateway. By indicating when shorter, lower-emission Arctic transits are feasible while limiting the risk of ice-related accidents in this sensitive polar environment, the framework also contributes to the safety and environmental sustainability of trans-Arctic shipping. Full article

Machine learning-based network intrusion detection systems are often optimized using aggregate accuracy, although operational security depends on the reliable detection of rare, high-impact attacks. This paper proposes a data-preserving intrusion detection framework that combines hybrid feature selection, heterogeneous ensemble learning and cost-sensitive optimization for imbalanced multi-class attack detection. The method first applies Mutual Information filtering and Recursive Feature Elimination to reduce the NSL-KDD feature space from 122 one-hot encoded attributes to 25 discriminative features. Four classifiers, Random Forest, XGBoost, Support Vector Machine and K-Nearest Neighbors, are evaluated individually, and a stacking ensemble is constructed using Logistic Regression as a meta-learner. Class imbalance is addressed by balanced class weighting rather than by synthetic oversampling, preserving the original minority-class observations. Experiments on the NSL-KDD benchmark show that the proposed cost-sensitive configuration improves rare attack recognition, most notably increasing U2R recall from 0.00% to 35.82% (24 of 67 test instances) for the stacking ensemble; this improvement, together with the accompanying weighted F1-score change from 0.7120 to 0.7214, is statistically significant under the Wilcoxon signed-rank test across repeated random seeds, and both values are reported with their variability rather than as single point estimates. SVM obtains the largest global gain, with a 7.06 percentage point improvement in weighted F1-score. The results show that cost-sensitive learning is a simple and practical mechanism for improving rare-attack visibility, but also reveal a remaining limitation for R2L detection, where feature overlap with Normal traffic remains substantial. The revised validation design explicitly includes direct resampling baselines, repeated-seed evaluation, statistical significance testing, feature-subset sensitivity analysis, and absolute true-positive counts for R2L and U2R in order to avoid overinterpreting marginal point-estimate gains. All experiments, including the resampling comparison, the component ablation, the feature-subset sensitivity analysis and the repeated-seed statistical evaluation, are executed on the complete KDDTrain+ training set of 125,973 instances under a single unified protocol, so that every reported per-class value refers to the same experimental setting. The revised study additionally reports probability-level evaluation for the primary model, including class-level PR-AUC, precision-recall curves and a U2R threshold and alert-budget analysis, and validates the framework externally on the UNSW-NB15 benchmark, where balanced class weighting raises the recall of the rarest categories (Worms, Shellcode, Backdoor) from near-zero baseline levels to 69–96% under an identical protocol. Full article

This paper proposes a method of fault detection in analog circuits which involves various steps, including selection of benchmark circuits, dataset preparation, signal decomposition, model training, and performance analysis. The main aim of this work is to provide solid performance even in noisy environments. Monte Carlo analysis is used to generate a synthetic dataset with 200 runs per fault class by introducing component tolerances and realistic faults. A multi-stage pipeline is proposed; it begins with resampling the signals and normalizing them, and then noise is added at different levels: 5 dB, 10 dB and 20 dB. Feature fusion is performed by combining time-, frequency-, and statistical-domain features. Statistical-domain features are extracted by applying Variational Mode Decomposition (VMD) to split them into four IMF levels, followed by the application of Continuous Wavelet Transform (CWT) for time–frequency-domain analysis. Support Vector Machine (SVM), Random Forest, and Gradient Boosting are used as base-level classification models. A stacking ensemble model is developed which uses Random Forest, Gradient Boosting, and Extra Trees as base learners and Logistic Regression as the meta-learner. Full article

Accurate large-scale crop mapping is fundamental to agricultural management. However, in Liaoning Province, undulating terrain and fragmented fields make fine crop classification challenging. In particular, corn and soybean have overlapping phenologies, which can lead to spectral and structural confusion in conventional optical–SAR feature spaces and limit mapping accuracy. This study proposes a fine crop mapping framework integrating optical phenotypic, microwave structural, and meteorological time-series features. To overcome the curse of dimensionality caused by high-dimensional heterogeneous data, an adaptive feature truncation mechanism based on the transition pattern of the marginal-gain curve was designed. Additionally, a pyramid multi-scale sliding window algorithm was constructed to optimize meteorological features, achieving dimensionality reduction and precise identification of phenologically sensitive windows. The results indicate that: (1) The multi-scale feature selection strategy effectively eliminates redundant variables and maximizes the inter-class discriminability of core features, significantly improving computational efficiency and classification performance. (2) High-frequency meteorological features provide key physiological constraints. Specifically, mid-May shortwave radiation, early October precipitation, and early August growing degree days constitute the core environmental–physiological features for distinguishing confused crops, helping to mitigate the spectral confusion of dryland crops. (3) Driven by the multi-source features, the Support Vector Machine (SVM) exhibits the optimal generalization robustness for processing high-dimensional structured data, yielding an overall classification accuracy of 91.80% and a Kappa coefficient of 0.8905. This framework provides a reliable methodological reference for high-precision crop monitoring in large-scale complex planting areas. Full article

To advance crop type mapping in support of global food and water security, this study compared three spectral configurations: (A) the full 60-band DLR Earth Sensing Imaging Spectrometer (DESIS) hyperspectral narrowband (HNB) dataset, (B) a 14-band subset of DESIS-derived HNBs aligned with the planned Landsat 10 (formerly Landsat Next) spectral configuration (400–1000 nm), and (C) DESIS-based simulations of Landsat 10 superspectral broadbands. The analysis was conducted in California’s Central Valley, hereafter referred to as “the Central Valley”, during the peak growing month of August. DESIS imagery from August 2021, 2022, and 2023 was used sequentially for model development, testing, and independent validation. Over these three years, DESIS provided extensive hyperspectral coverage of much of the 4 million hectares in the Central Valley’s. Analyses were performed on Google Earth Engine using two pixel-based supervised classifiers, Random Forest (RF) and Support Vector Machine (SVM), to differentiate three major crop classes: row crops, grapes and tree crops, and winter wheat/fallow/other. The highest overall accuracy (86%) was achieved using SVM in combination with either the full DESIS hyperspectral dataset or the 14 DESIS narrowbands corresponding to Landsat 10. This finding aligns with earlier studies showing a small number of strategically positioned narrowbands can be optimal for crop type classification. Use of the narrowband datasets resulted in substantially higher accuracy (overall accuracy of 86%) compared to the simulated Landsat 10 broadbands (overall accuracy of 75%), supporting previous studies highlighting the utility of narrowbands. Despite the high accuracy using August imagery, the study indicates more granular crop type classification will require multi-temporal observations spanning the full phenological cycle (June–October), especially for a large number of crop classes. Acquiring task-based hyperspectral imagery over such large areas throughout the growing season remains operationally challenging. In contrast, Landsat 10 superspectral imagery could provide routine coverage across seasons and years that is practical and scalable for future large area crop type mapping and agricultural monitoring. Full article

Grasslands represent one of the most extensive terrestrial biomes globally, covering approximately one-third of the Earth’s land surface, yet they are increasingly threatened by land-use change and overgrazing, underscoring the need for reliable monitoring approaches. This study compares the performance of machine learning and deep learning algorithms for grassland mapping using multi-source remote sensing data derived from Sentinel-1, Sentinel-2, and terrain variables. The research was conducted in Mpumalanga Province, South Africa, a heterogeneous landscape comprising lowland savannas, high-altitude grasslands, escarpments, and riverine wetlands. Random Forest (RF) and Support Vector Machine (SVM) classifiers were implemented in Google Earth Engine using fused satellite and terrain datasets with field-collected samples for training and validation, while a One-Dimensional Convolutional Neural Network (1D-CNN) was developed in Python 3.13.5 using the same inputs. Results demonstrate that integrating multi-source data improves classification accuracy, with radar-based features contributing the most. RF achieved the highest performance, with an overall accuracy of 97.7% and grass-class precision, recall, and F1-score exceeding 0.97, closely followed by the 1D-CNN with 91% overall accuracy and complete grass detection. In contrast, SVM performed notably lower with an overall accuracy of 80,8%. These findings highlight the effectiveness of advanced learning approaches for grassland mapping and support their application in ecological restoration and environmental management. Full article

The Chang 8 Member of the Jinghe Oilfield in the Ordos Basin is a low-porosity, ultra-low-permeability reservoir with many faults and fractures, complex structures, and strong heterogeneity. Conventional logging curves do not clearly distinguish among different structural units, making it difficult to identify the internal structures of fault-fracture reservoirs. Current methods mainly use logging curves and rock mechanical parameters. In these reservoirs, experiments are costly, numerical simulations take a long time, and identification is often inefficient. To improve identification accuracy and efficiency, this study developed a two-layer Stacking ensemble model for the Chang 8 Member. The dataset was derived from conventional well-log data from five wells in the Chang 8 Member and contained 816 labelled depth samples. Among them, 569 original samples from wells A1, A2, and A3 were used for model development, while 247 samples from wells JH55P10 and JH2301H were reserved for independent well-level validation. In the first layer, a support vector machine (SVM), XGBoost, and a random forest (RF) were used as the base learners. The hyperparameters of the base learners were optimized using grid search and K-fold cross-validation. In the second layer, multinomial logistic regression was used as the meta-learner to integrate the class-probability outputs of the base learners and generate the final predictions. Individual models showed limitations in distinguishing the three internal structural units of fault-fracture reservoirs. By integrating the complementary outputs of the base learners, the Stacking model achieved an overall accuracy of 0.89, exceeding the accuracies of the individual models on the internal hold-out test set. The results indicate that the proposed framework can improve the accuracy and class balance of multi-class identification on the present dataset and provide a practical approach for the detailed evaluation of internal structural units in low-porosity, low-permeability fault-fracture reservoirs. Full article

Accurate fault location is essential for improving the reliability and service restoration capability of distribution networks. With the increasing penetration of distributed generation, power electronic devices, and flexible loads, fault transient signals become increasingly nonlinear and nonstationary, posing challenges to conventional impedance-based, traveling-wave-based, and feature-engineering-based methods. To improve transient fault feature representation, this paper proposes a wavelet-embedded residual attention convolutional neural network (CNN) for distribution network fault location. The task is formulated as a multi-class classification problem, in which each predefined line section is treated as a candidate fault location class. The proposed method embeds discrete wavelet decomposition into the convolutional feature extraction process, enabling low-frequency trend components and high-frequency transient components to be jointly represented and fused by subsequent trainable network modules. Residual connections improve deep feature propagation, and an attention mechanism enhances fault-sensitive representations. Simulation studies on the IEEE 33-bus distribution system show that the proposed method outperforms multi-layer perceptron (MLP), support vector machine (SVM), standard CNN, ResNet, and Attention-CNN, achieving 98.27% accuracy and a 98.33% F1-score. The class-wise results and robustness tests under different transition resistances, noise levels, and fault types further verify the effectiveness and adaptability of the proposed method. Full article