K-nearest Neighbor Research Articles

A multiscale molecular structural neural network for molecular property prediction.

Molecular Property Prediction (MPP) is a fundamental task in important research fields such as chemistry, materials, biology, and medicine, where traditional computational chemistry methods based on quantum mechanics often consume substantial time and computing power. In recent years, machine learning has been increasingly used in computational chemistry, in which graph neural networks have shown good performance in molecular property prediction tasks, but they have some limitations in terms of generalizability, interpretability, and certainty. In order to address the above challenges, a Multiscale Molecular Structural Neural Network (MMSNet) is proposed in this paper, which obtains rich multiscale molecular representations through the information fusion between bonded and non-bonded "message passing" structures at the atomic scale and spatial feature information "encoder-decoder" structures at the molecular scale; a multi-level attention mechanism is introduced on the basis of theoretical analysis of molecular mechanics in order to enhance the model's interpretability; the prediction results of MMSNet are used as label values and clustered in the molecular library by the K-NN (K-Nearest Neighbors) algorithm to reverse match the spatial structure of the molecules, and the certainty of the model is quantified by comparing virtual screening results across different K-values. Experimental results in the authoritative small molecule dataset QM9 and the macromolecular protein database PDBbind demonstrate that MMSNet has optimal prediction accuracy, model complexity, and generalizability compared with more than ten existing state-of-the-art (SOTA) models in a variety of different types of prediction tasks; it has a great potential for downstream tasks such as chemical research, drug discovery, and material design.

Detecting B-cell lymphoma-6 overexpression status in primary central nervous system lymphoma using multiparametric MRI-based machine learning.

In primary central nervous system lymphoma (PCNSL), B-cell lymphoma-6 (BCL-6) is an unfavorable prognostic biomarker. We aim to non-invasively detect BCL-6 overexpression in PCNSL patients using multiparametric MRI and machine learning techniques. 65 patients (101 lesions) with primary central nervous system lymphoma (PCNSL) diagnosed from January 2013 to July 2023, and all patients were randomly divided into a training set and a validation set according to a ratio of 8 to 2. ADC map derived from DWI (b = 0/1000 s/mm2), fast spin echo T2WI, T2FLAIR, were collected at 3.0 T. A total of 2234 radiomics features from the tumor segmentation area were extracted and LASSO were used to select features. Logistic regression (LR), Naive bayes (NB), Support vector machine (SVM), K-nearest Neighbor, (KNN) and Multilayer Perceptron (MLP), were used for machine learning, and sensitivity, specificity, accuracy F1-score, and area under the curve (AUC) was used to evaluate the detection performance of five classifiers, 6 groups with combinations of different sequences were fitted to 5 classifiers, and optimal classifier was obtained. BCL-6 status could be identified to varying degrees with 30 models based on radiomics, and model performance could be improved by combining different sequences and classifiers. Support vector machine (SVM) combined with three sequence group had the largest AUC (0.95) in training set and satisfactory AUC (0.87) in validation set. Multiparametric MRI based machine learning is promising in detecting BCL-6 overexpression.

Feasibility study on fingerprinting organic and conventional mango fruits, chips, and juice using portable near-infrared spectroscopy.

This research examined the distinction between organic and conventional mango fruits, chips, and juice using portable near-infrared (NIR) spectroscopy. A comprehensive analysis was conducted on a sample of 100 mangoes (comprising 50 organic and 50 conventional) utilising a portable NIR spectrometer that spans a wavelength range from 900 to 1700 nm. The mangoes were assessed in their entirety and their juice and chip forms. The spectral data underwent pre-processing through methodologies such as multiplicative scatter correction (MSC), standard normal variate (SNV), and derivatives to enhance the precision of the models. Principal component analysis (PCA) and various multivariate classification algorithms, including linear discriminant analysis (LDA), random forest (RF), k-nearest neighbors (kNN), and partial least squares discriminant analysis (PLSDA), were utilised to categorise the samples effectively. The findings indicated that the random forest method and specific pre-processing techniques achieved the highest classification accuracy for distinguishing organic and conventional mango products. For mango fruit and chips, it achieved 88.76% and 77.98% accuracy, respectively, when pre-processed using the second derivative, while for juice, it achieved 87.53% accuracy without pre-processing. This investigation demonstrates the efficacy of portable NIR spectroscopy as a dependable and non-invasive method for verifying organic mango products, thereby enhancing the integrity of food labelling and fostering consumer confidence.

Monkeypox diagnosis based on probabilistic K-nearest neighbors (PKNN) algorithm.

Although it is not a new illness and has been around since the previous century, monkeypox later resurgence is fraught with difficulties. This study presents a novel approach of diagnosing monkeypox using artificial intelligence, which is called Effective Monkeypox Diagnosis Strategy (EMDS). The proposed EMDS is established through two sequential stages, namely; (i) Pre-Processing Phase (PP) and (ii) Monkeypox Diagnosing phase (MDP). During PP the input image dataset is prepared through three processes, which are; feature extraction, feature selection, and anomaly rejection, while the actual diagnosis performed in the MDP. Features are extracted from input images using GoogleNet as an effective pre-trained deep learning model, while Leopard Seal Optimization (LSO) is employed to select the most instructive features. The major contribution of this paper is focused in two issues, which are; (i) introducing a new methodology for rejecting anomalies from the input image dataset based on interquartile range (IQR), and (ii) proposing a new instance of K-Nearest Neighbor classifier for monkeypox diagnosis, which is called; Probabilistic K-Nearest Neighbors (PKNN) Algorithm. The proposed PKNN combines evidence from distance based traditional KNN as well as the Naïve probabilistic theorem used by Naïve Bayes (NB) algorithm in an integrated way. Numerous experiments have been conducted considering the proposed EMDS as well as recent competitive strategies on two public monkeypox datasets, which are; the Monkeypox Skin Image and Lesion Datasets (MSID and MSLD, respectively). Initially, the performance of the basic contributions of the proposed EMDS, which are the outlier rejection methodology (ORM) and PKNN, are evaluated individually. Then, EMSD as a whole is evaluated. Moreover, an ablation study has also been conducted to evaluate the effect of ORM and PKNN on the performance of EMSD. Based on the experimental results, it is shown that EMDS outperforms recent monkeypox identification strategies as it achieves 99% diagnosis accuracy. Moreover, it indicates the maximum precision and recall with the minimum diagnosis time.

Erlang Spectrogram and Residual Network-Based Features for Fake Audio Detection

Advancements in deep learning and other approaches are making synthetic speech more natural-sounding. These advancements enable people to train a speech synthesizer with a target voice, resulting in a model that can imitate someone's voice with excellent accuracy. To overcome these challenges, in this work, we are presenting a fusion of Spectrogram and Deep Convolutional Neural Network (DeepCNN) for detecting fake audio. The work has been implemented in three phases: Spectrogram generation, feature extraction & reduction, and classification. In the first step, Erlang spectrograms have been generated using the proposed Erlang filter bank. These generated spectrograms are then passed to the Residual Network (ResNet11) for feature extraction, and later these extracted features are fed to the Linear Discriminant Analysis (LDA) for dimensionality reduction. In the last phase, LDA-optimized features were used to train the four machine learning classifiers, one by one, eXtreme Gradient Boosting (XGboost), Naïve Bayes (NB), KNN (K-Nearest Neighbor), and Random Forest (RF). The proposed work has been evaluated on the complete Fake or Real (FoR) dataset for detecting fake audio in different scenarios. The combination of Erlang spectrogram-ResNet11-LDA with XGBoost has achieved a minimum 0% and 1.4% EER for FoR original and FoR 2sec.

Determinants Shaping Human Preference for Thermal Springs in Palaeolithic Europe and Asia Minor

ABSTRACT This study explores Palaeolithic human occupation at thermal springs in Europe and Asia Minor by reconstructing palaeoecological conditions for 12 thermal spring sites and 97 control Palaeolithic sites using machine learning and dimensionality reduction techniques. K-nearest neighbours analysis indicates that thermal spring sites were occupied under hot, dry Mediterranean, oceanic, or humid continental climates. Age-specific δ18O data suggest these sites were mostly used during cold or moderately cold global climatic conditions.Principal Component Analysis revealed that thermal spring sites experienced drier climates with higher diurnal and annual temperature ranges than control sites. Two distinct clusters emerged based on reconstructed palaeoenvironmental conditions: a temperate cluster, subdivided into Mediterranean and oceanic subclusters, and a humid continental cluster. Linear Discriminant Analysis clearly separated thermal spring sites from non-thermal ones.Ensemble algorithms identified temperature annual range, annual precipitation, and clothing coverage as key differentiators between thermal spring and control sites, depending on paleoclimate. This analysis highlights the varied motivations for Palaeolithic occupations at thermal springs across diverse European and Asia Minor environments.

Integrating Retinal Segmentation Metrics with Machine Learning for Predictions from Mouse SD-OCT Scans

Purpose This study aimed to initially test whether machine learning approaches could categorically predict two simple biological features, mouse age and mouse species, using the retinal segmentation metrics. Methods The retinal layer thickness data obtained from C57BL/6 and DBA/2J mice were processed for machine learning after segmenting mouse retinal SD-OCT scans. Twenty-two models were trained to predict the mouse groups. The best neural network model was optimized for better outcomes. Prediction accuracy, the area under the curve, sensitivity, specificity, precision, and F-1 score values were obtained. Results The Wilcoxon Signed-Rank test provided significantly higher validation accuracy for neural networks than decision trees, discriminant analysis, support vector machines, and k-nearest neighbor classifiers (p = 0.005 for all). For C57BL/6-DBA/2J classification, a mean validation accuracy of 88.11 ± 3.92% (95% CI: 86.99-89.22) was achieved for the neural network when the optimized neural network had 92.31% final test accuracy with an area under the curve value of 0.9762, 94.44% sensitivity, 90.48% specificity, 89.47% precision, and 0.92 F-1 score. The optimized neural network model for age group differentiation had a final test accuracy of 82.05% with a 0.9064 area under the curve value, 77.27% sensitivity, 88.24% specificity, 89.47% precision, and 0.83 F-1 score. Conclusions These findings validate that machine learning, using segmentation metrics instead of images, can effectively analyze retinal OCT scans in mice for categorical predictions in experimental models. Expanding this approach with additional features, including histopathological and functional correlations, is expected to improve the prediction power further, promising valuable applications to predict more complex outcomes in experimental and clinical studies.

Essential Oils as Antimicrobials against Acinetobacter baumannii: Experimental and Literature Data to Definite Predictive Quantitative Composition-Activity Relationship Models Using Machine Learning Algorithms.

Essential oils (EOs) exhibit a broad spectrum of biological activities; however, their clinical application is hindered by challenges, such as variability in chemical composition and chemical/physical instability. A critical limitation is the lack of chemical consistency across EO samples, which impedes standardization. Despite this, evidence suggests that EOs with differing chemical profiles often display similar (micro)biological activities, raising the possibility of standardizing EOs based on their biological effects rather than their chemical composition. This study explored the relationship between EO chemical composition and antibacterial activity against carbapenem-resistant Acinetobacter baumannii. A dataset comprising 82 EOs with known minimal inhibitory concentration values was compiled using both experimental results and literature data sourced from the AI4EssOil database (https://www.ai4essoil.com). Machine learning classification algorithms including Support Vector Machines, Random Forest, Gradient Boosting, Decision Trees, and K-Nearest Neighbors were employed to generate quantitative composition-activity relationship models. Model performance was assessed using internal and external prediction accuracy metrics with the Matthews correlation coefficient as the primary evaluation metrics. Features importance analysis, based on the Skater methodology, identified key chemical components influencing EO activity. The single chemical components limonene, eucalyptol, alpha-pinene, linalool, beta-caryophyllene, nerol, beta-pinene, neral, and carvacrol were highlighted as critical to biological efficacy. The predictive capacity of the ML models was validated against a test set of freshly extracted and chemically characterized EOs. The models demonstrated a 91% prediction accuracy for new EO samples, and a strong correlation was observed between predicted features importance and experimental inhibitory values for six selected pure compounds (limonene, eucalyptol, alpha-pinene, linalool, carvacrol, and thymol). Additionally, the machine learning approach was extended to cytotoxicity data from 3T3-Swiss fibroblasts for 61 EOs. The analysis revealed the potential to design EOs with both high antibacterial activity and low cytotoxicity through blending or selective enrichment with identified key components. These findings pave the way for biologically standardized EOs, enabling their rational design and optimization for clinical applications.

Development of a machine learning tool to predict deep inspiration breath hold requirement for locoregional right-sided breast radiation therapy patients

Background and purpose. This study presents machine learning (ML) models that predict if deep inspiration breath hold (DIBH) is needed based on lung dose in right-sided breast cancer patients during the initial computed tomography (CT) appointment.Materials and methods. Anatomic distances were extracted from a single-institution dataset of free breathing (FB) CT scans from locoregional right-sided breast cancer patients. Models were developed using combinations of anatomic distances and ML classification algorithms (gradient boosting, k-nearest neighbors, logistic regression, random forest, and support vector machine) and optimized over 100 iterations using stratified 5-fold cross-validation. Models were grouped by the number of anatomic distances used during development; those with the highest validation accuracy were selected as final models. Final models were compared based on their predictive ability, measurement collection efficiency, and robustness to simulated user error during measurement collection.Results. This retrospective study included 238 patients treated between 2016 and 2021. Model development ended once eight anatomic distances were included, and the validation accuracy plateaued. The best performing model used logistic regression with four anatomic distances achieving 80.5% average testing accuracy, with minimal false negatives and positives (<27%). The anatomic distances required for prediction were collected within 3 min and were robust to simulated user error during measurement collection, changing accuracy by <5%.Conclusion. Our logistic regression model using four anatomic distances provided the best balance between efficiency, robustness, and ability to predict if DIBH was needed for locoregional right-sided breast cancer patients.

Using near-infrared hyperspectral imaging combined with machine learning to predict the components and the origin of Radix Paeoniae Rubra.

The efficacy and safety of drugs are closely related to the geographical origin and quality of the raw materials. This study focuses on using near-infrared hyperspectral imaging (NIR-HSI) combined with machine learning algorithms to construct content prediction models and origin identification models to predict the components and origin of Radix Paeoniae Rubra (RPR). These models are quick, non-destructive, and accurate for assessing both component content and origin. Spectral data were preprocessed using multiple scattering correction (MSC), Savitzky-Golay smoothing (S-G), and standard normal variate (SNV). Content prediction models for paeoniflorin were developed using principal component regression (PCR), partial least squares regression (PLSR), and ridge regression (RR). Classification models for origin identification utilized support vector machine (SVM), K-nearest neighbor (KNN), and random forest (RF). The SNV-RR model achieved a determination coefficient of 0.8943, while the SNV-SVM model achieved an accuracy of 0.9790. Meanwhile, two feature selection methods were used to further simplify the prediction model while ensuring accuracy, in order to improve the detection efficiency in practical applications. This research demonstrates the feasibility of combining NIR-HSI with machine learning for quality analysis of RPR, providing a theoretical basis for promoting hyperspectral imaging technology in the food and pharmaceutical sectors.

AI Machine Learning-Based Diabetes Prediction in Older Adults in South Korea: Cross-Sectional Analysis.

Diabetes is prevalent in older adults, and machine learning algorithms could help predict diabetes in this population. This study determined diabetes risk factors among older adults aged ≥60 years using machine learning algorithms and selected an optimized prediction model. This cross-sectional study was conducted on 3084 older adults aged ≥60 years in Seoul from January to November 2023. Data were collected using a mobile app (Gosufit) that measured depression, stress, anxiety, basal metabolic rate, oxygen saturation, heart rate, and average daily step count. Health coordinators recorded data on diabetes, hypertension, hyperlipidemia, chronic obstructive pulmonary disease, percent body fat, and percent muscle. The presence of diabetes was the target variable, with various health indicators as predictors. Machine learning algorithms, including random forest, gradient boosting model, light gradient boosting model, extreme gradient boosting model, and k-nearest neighbors, were employed for analysis. The dataset was split into 70% training and 30% testing sets. Model performance was evaluated using accuracy, precision, recall, F1 score, and area under the curve (AUC). Shapley additive explanations (SHAPs) were used for model interpretability. Significant predictors of diabetes included hypertension (χ²1=197.294; P<.001), hyperlipidemia (χ²1=47.671; P<.001), age (mean: diabetes group 72.66 years vs nondiabetes group 71.81 years), stress (mean: diabetes group 42.68 vs nondiabetes group 41.47; t3082=-2.858; P=.004), and heart rate (mean: diabetes group 75.05 beats/min vs nondiabetes group 73.14 beats/min; t3082=-7.948; P<.001). The extreme gradient boosting model (XGBM) demonstrated the best performance, with an accuracy of 84.88%, precision of 77.92%, recall of 66.91%, F1 score of 72.00, and AUC of 0.7957. The SHAP analysis of the top-performing XGBM revealed key predictors for diabetes: hypertension, age, percent body fat, heart rate, hyperlipidemia, basal metabolic rate, stress, and oxygen saturation. Hypertension strongly increased diabetes risk, while advanced age and elevated stress levels also showed significant associations. Hyperlipidemia and higher heart rates further heightened diabetes probability. These results highlight the importance and directional impact of specific features in predicting diabetes, providing valuable insights for risk stratification and targeted interventions. This study focused on modifiable risk factors, providing crucial data for establishing a system for the automated collection of health information and lifelog data from older adults using digital devices at service facilities.

Time series forecasting of bed occupancy in mental health facilities in India using machine learning

Machine learning models are vital for forecasting and optimizing healthcare parameters, especially in the context of rising mental health issues in India and globally. With increasing demand for mental health services, effective resource management, like bed occupancy forecasting, is crucial to ensure proper patient care and reduce the burden on healthcare facilities. This study applies six machine learning models, namely Support Vector Regression, eXtreme Gradient Boosting, Random Forest, K-Nearest Neighbors, Gradient Boosting, and Decision Tree, to forecast weekly bed occupancy of the second largest mental hospital in India, using data from 2008 to 2024. Accuracy of models were evaluated using Mean Absolute Percentage Error, and Diebold–Mariano test for assessing differences in predictive performance. Further, we forecast the bed occupancy, providing crucial insights for healthcare administrators in capacity planning and resource allocation, supporting data-driven decisions and enhancing the quality of mental health services in India.

Wood-waste classification using interpretable machine learning

ABSTRACT Determining and classifying the contents of wood wastes, which possess an extremely heterogeneous structure, is crucial for their optimal utilization. The heterogeneous nature of waste wood poses significant challenges for its separation and classification. Also, the classical chemometric methods are not ergonomic in terms of time and cost. For this reason, this study aimed to develop an efficient ML-based decision support system (DSS) for accurate classification of waste wood in addition to being an ergonomic approach versus chemometric methods. To develop ML-based DSS, firstly absorbance values at 650–4000 cm−1 wavelength were measured using FTIR-ATR device for 200 wood-waste samples. Absorbance values at 52 key wavelengths, identified as the most effective for characterizing the samples, were used as input features to estimate ML classification models using artificial neural networks (ANN), random forest (RF), ensemble learning, multiple logistic regression (MLR), discriminant analysis, Naive Bayes, K-nearest neighbour (KNN), and support vector machines (SVM). The analysis results indicated that the developed DSS achieved quite a high classification accuracy rates, approaching 100%. As a result, this kind of ML-based DSS can be used as a practical and efficient tool to determine and classify the contents of wood wastes.

Optimized machine learning approaches to combine surface-enhanced Raman scattering and infrared data for trace detection of xylazine in illicit opioids.

Infrared absorption spectroscopy and surface-enhanced Raman spectroscopy were integrated into three data fusion strategies-hybrid (concatenated spectra), mid-level (extracted features from both datasets) and high-level (fusion of predictions from both models)-to enhance the predictive accuracy for xylazine detection in illicit opioid samples. Three chemometric approaches-random forest, support vector machine, and k-nearest neighbor algorithms-were employed and optimized using a 5-fold cross-validation grid search for all fusion strategies. Validation results identified the random forest classifier as the optimal model for all fusion strategies, achieving high sensitivity (88% for hybrid, 92% for mid-level, and 96% for high-level) and specificity (88% for hybrid, mid-level, and high-level). The enhanced performance of the high-level fusion approach (F1 score of 92%) is demonstrated, effectively leveraging the surface-enhanced Raman data with a 90% voting weight, without compromising prediction accuracy (92%) when combined with infrared spectral data. This highlights the viability of a multi-instrument approach using data fusion and random forest classification to improve the detection of various components in complex opioid samples in a point-of-care setting.

MVGNN-PPIS: A novel multi-view graph neural network for protein-protein interaction sites prediction based on Alphafold3-predicted structures and transfer learning.

Protein-protein interactions (PPI) are crucial for understanding numerous biological processes and pathogenic mechanisms. Identifying interaction sites is essential for biomedical research and targeted drug development. Compared to experimental methods, accurate computational approaches for protein-protein interaction sites (PPIS) prediction can save significant time and costs. In this study, we propose a novel model named MVGNN-PPIS. To the best of our knowledge, it is the first to utilize predicted structures generated by AlphaFold3, and combined with transfer learning techniques, for predicting PPIS. This approach addresses the limitations of traditional methods that depend on native protein structures and multiple sequence alignments (MSA). Additionally, we introduced a multi-view graph framework based on two types of graph structures: the k-nearest neighbor graph and the adjacency matrix. By alternately employing a Graph Transformer and Graph Convolutional Networks (GCN) to aggregate node information, this framework effectively captures both local and global dependencies of each residue in the predicted structures, thereby significantly enhancing the model's sensitivity to binding sites. This framework further integrates direction, distances and angular information between the 3D coordinates of side-chain atom centroids to construct a relative coordinate system, generating enhanced edge features that ensure the model's equivariance to molecular translations and rotations in space. During training, the Focal Loss function is employed to effectively address the class imbalance in the dataset. Experimental results demonstrate that MVGNN outperforms the current state-of-the-art methods across multiple PPIS benchmark datasets. To further validate the model's generalization capability, we extended MVGNN to the domain of predicting protein-nucleic acid interaction sites, where it also achieved superior performance.

K-Nearest Neighbour Estimation of the Conditional Set-Indexed Empirical Process for Functional Data: Asymptotic Properties

k-Nearest Neighbour Estimation of the Conditional Set-Indexed Empirical Process for Functional Data: Asymptotic Properties

Robust fault detection and classification in power transmission lines via ensemble machine learning models

Transmission lines are vital for delivering electricity over long distances, yet they face reliability challenges due to faults that can disrupt power supply and pose safety risks. This research introduces a novel approach for fault detection and classification by analyzing voltage and current patterns across transmission line phases. Leveraging a comprehensive dataset of diverse fault scenarios, various machine learning algorithms—including Random Forest (RF), K-Nearest Neighbors (KNN), and Long Short-Term Memory (LSTM) networks—are evaluated. An ensemble methodology, RF-LSTM Tuned KNN, is proposed to enhance detection accuracy and robustness. Results indicate that RF-LSTM Tuned KNN achieves a remarkable accuracy of 99.96% on a multi-label dataset, outperforming RF (97.50%) and KNN (96.55%). In binary classification, KNN attains the highest accuracy of 99.85%, closely followed by RF at 99.72%. This methodology provides significant advancements in fault detection capabilities, offering valuable insights for improving grid reliability and stability, and ensuring a more resilient power supply.

Machine learning models for easily obtainable descriptors of the electrocatalytic properties of Ag-Pd-Ir nanoalloys toward the formate oxidation reaction.

Direct formate fuel cells (DFFCs) have received increasing attention due to their environmentally benign and highly safe characteristics. However, the absence of highly active electrocatalysts for the formate oxidation reaction (FOR) restricts their widespread application. Currently, the design of FOR catalysts, which relies on experimental trial-and-error and high-throughput DFT calculations, is costly and time-consuming. In this study, based on a DFT dataset of FOR overpotentials for 137 Ag-Pd-Ir nanoalloy catalysts, six machine learning (ML) models were trained, where the K-nearest neighbors (KNN) model demonstrated the best performance, with an R2 value of 0.94, an MAE value of 0.041 V and an RMSE value of 0.050 V. Using the KNN model, six optimal catalysts with an overpotential of 0.48 V were screened from 310 candidate catalysts, with an MAE value as low as 0.004 V compared to the DFT results, proving the accuracy of the ML model. This work provides a novel strategy to accelerate the design of high-performance catalysts.

Modified kernel global-local marginal fisher analysis for rolling bearing feature extraction

Abstract The vibration signals of rolling bearings usually exhibit high-dimensional, nonlinear, and non-Gaussian distribution characteristics due to long-term operation under complex working conditions. Therefore, we proposed a novel algorithm named Modified Kernel Global-Local Marginal Fisher Analysis (MKGLMFA) for bearing feature extraction and dimensionality reduction. The proposed MKGLMFA algorithm introduces the kernel function to map data into a high-dimensional space to represent data nonlinearly first. It enhances the within-class compactness and between-class dispersibility by considering spatial relationships and label information when constructing adjacency graphs and simultaneously exploits the local and global geometry of data. Furthermore, a bearing fault diagnosis approach is presented based on MKGLMFA. It first processes the original vibration signals through MKGLMFA to obtain low-dimensional manifold features. Then these characteristics were input into the K-nearest neighbor (KNN) classifier to achieve fault pattern recognition. The superiority of the proposed MKGLMFA algorithm in feature extraction is verified in comparison with some existing state-of-the-art machine learning methods on three rolling bearings datasets. And the subsequent classification diagnosis experiments indicate the effectiveness and high efficiency of the newly raised MKGLMFA algorithm. In comparison with the representative diagnosis methods, the proposed method can extract more sensitive discriminant features, and the classification accuracy of diagnosis is significantly improved in consequence.

Analysis of the Style Characteristics of Regional Folk Songs and Music Classification Algorithms

Regional folk songs have a rich history and are filled with cultural values. In this paper, first, the style characteristics of regional folk songs are briefly introduced. Using four regional folk songs from the northwest, northeast, southwest, and Hakka as examples, time domain, frequency domain, and mel-frequency cepstral coefficient (MFCC) features were extracted. Finally, the bidirectional long short-term memory (BiLSTM)-based music classification algorithm is used to realize the classification of folk songs from different regions. It was found that using the time-frequency domain + MFCC as features produced better results in music classification than using only the time-frequency domain or only MFCC features. The BiLSTM algorithm achieved an accuracy of 0.8339 and an F1 value of 0.8201 for the 10 s fragment set, both of which were better than those of the K-nearest neighbor, support vector machine, and other classification algorithms. The results show that the approach used in this study to categorize regional folk songs is reliable and that it can be applied to real folk songs.