Hierarchical Density-Based Spatial Clustering Of Applications With Noise Research Articles

Computational screening of sodium solid electrolytes through unsupervised learning

All-solid-state Na-ion batteries have emerged as alternatives to all-solid-state Li-ion batteries owing to the global abundance of Na element. However, finding a commercially viable Na-ion solid-state electrolyte (SSE) remains challenging due to the relatively poor understanding of the structures effective for conduction compared to those for Li-ion SSE. In this study, we develop a screening framework based on an unsupervised machine learning technique to characterize Na-ion SSEs according to their lattice structures. Specifically, we evaluate feature vectors encoding 180 structural properties for 12,670 materials containing Na ions. Subsequently, the resulting feature vectors are clustered using hierarchical density-based spatial clustering of applications with noise (HDBSCAN), leading to the discovery of 12 groups including those with experimentally proven Na-ion superionic conductors such as NASICONs and sodium chalcogenides. Post hoc analysis of these clusters reveals that the groups with high conductivity share similar characteristics, including the existence of ion channels for Na ions and the weak interactions between Na ions and the proximate atoms. Ab initio molecular dynamics simulations confirm that the promising groups exhibit exceptional ion diffusivity compared to other groups. By employing decision tree classifiers trained to screen promising groups, we demonstrate the rapid assessment of the potential of a given material. Finally, we offer perspectives and insights for the development of novel Na-ion SSEs for all-solid-state Na-ion batteries.

Development and Application of an Advanced Automatic Identification System (AIS)-Based Ship Trajectory Extraction Framework for Maritime Traffic Analysis

This study addresses the challenges of maritime traffic management in the western waters of Taiwan, a region characterized by substantial commercial shipping activity and ongoing environmental development. Using 2023 Automatic Identification System (AIS) data, this study develops a robust feature extraction framework involving data cleaning, anomaly trajectory point detection, trajectory compression, and advanced processing techniques. Dynamic Time Warping (DTW) and the Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithms are applied to cluster the trajectory data, revealing 16 distinct maritime traffic patterns, key navigation routes, and intersections. The findings provide fresh perspectives on analyzing maritime traffic, identifying high-risk areas, and informing safety and spatial planning. In practical applications, the results help navigators optimize route planning, improve resource allocation for maritime authorities, and inform the development of infrastructure and navigational aids. Furthermore, these outcomes are essential for detecting abnormal ship behavior, and they highlight the potential of route extraction in maritime surveillance.

Obstacle avoidance control of UGV based on adaptive-dynamic control barrier function in unstructured terrain

Abstract The widely used model predictive control of discrete-time control barrier functions (MPC-CBF) has difficulties in obstacle avoidance for unmanned ground vehicles (UGVs) in complex terrain. To address this problem, we propose adaptive dynamic control barrier functions (AD-CBF). AD-CBF is able to adaptively select an extended class of functions of CBF to optimize the feasibility and flexibility of obstacle avoidance behaviors based on the relative positions of the UGV and the obstacle, which in turn improves the obstacle avoidance speed and safety of the MPC algorithm when integrated with MPC. The algorithmic constraints of the CBF employ hierarchical density-based spatial clustering of applications with noise (HDBSCAN) for parameterization of dynamic obstacle information and unscaled Kalman filter (UKF) for trajectory prediction. Through simulations and practical experiments, we demonstrate the effectiveness of the AD-CBF-MPC algorithm in planning optimal obstacle avoidance paths in dynamic environments, overcoming the limitations of the point-by-point feasibility of MPC-CBF.

Supervised star, galaxy, and QSO classification with sharpened dimensionality reduction

We explored the use of broadband colors to classify stars, galaxies, and quasi-stellar objects (QSOs). Specifically, we applied sharpened dimensionality reduction (SDR)-aided classification to this problem, with the aim of enhancing cluster separation in the projections of high-dimensional data clusters to allow for better classification performance and more informative projections. The main objective of this work was to apply SDR to large sets of broadband colors derived from the CPz catalog to obtain projections with clusters of star, galaxy, and QSO data that exhibit a high degree of separation. The SDR method achieves this by combining density-based clustering with conventional dimensionality-reduction techniques. To make SDR scalable and have the ability to project samples using the earlier-computed projection, we used a deep neural network trained to reproduce the SDR projections. Subsequently classification was done by applying a $k$-nearest neighbors ($k$-NN) classifier to the sharpened projections. Based on a qualitative and quantitative analysis of the embeddings produced by SDR, we find that SDR consistently produces accurate projections with a high degree of cluster separation. A number of projection performance metrics are used to evaluate this separation, including the trustworthiness, continuity, Shepard goodness, and distribution consistency metrics. Using the $k$-NN classifier and consolidating the results of various data sets, we obtain precisions of 99.7<!PCT!>, 98.9<!PCT!>, and 98.5<!PCT!> for classifying stars, galaxies, and QSOs, respectively. Furthermore, we achieve completenesses of 97.8<!PCT!>, 99.3<!PCT!>, and 86.8<!PCT!>, respectively. In addition to classification, we explore the structure of the embeddings produced by SDR by cross-matching with data from Gaia DR3, Galaxy Zoo 1, and a catalog of specific star formation rates, stellar masses, and dust luminosities. We discover that the embeddings reveal astrophysical information, which allows one to understand the structure of the high-dimensional broadband color data in greater detail. We find that SDR-aided star, galaxy, and QSO classification performs comparably to another unsupervised learning method using hierarchical density-based spatial clustering of applications with noise (HDBSCAN) but offers advantages in terms of scalability and interpretability. Furthermore, it outperforms traditional color selection methods in terms of QSO classification performance. Overall, we demonstrate the potential of SDR-aided classification to provide an accurate and physically insightful classification of astronomical objects based on their broadband colors.

An integrated framework for flood disaster information extraction and analysis leveraging social media data: A case study of the Shouguang flood in China

Flood disasters cause significant casualties and economic losses annually worldwide. During disasters, accurate and timely information is crucial for disaster management. However, remote sensing cannot balance temporal and spatial resolution, and the coverage of specialized equipment is limited, making continuous monitoring challenging. Real-time disaster-related information shared by social media users offers new possibilities for monitoring. We propose a framework for extracting and analyzing flood information from social media, validated through the 2018 Shouguang flood in China. This framework innovatively combines deep learning techniques and regular expression matching techniques to automatically extract key flood-related information from Weibo textual data, such as problems, floodings, needs, rescues, and measures, achieving an accuracy of 83 %, surpassing traditional models like the Biterm Topic Model (BTM). In the spatiotemporal analysis of the disaster, our research identifies critical time points during the disaster through quantitative analysis of the information and explores the spatial distribution of calls for help using Kernel Density Estimation (KDE), followed by identifying the core affected areas using the Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm. For semantic analysis, we adopt the Latent Dirichlet Allocation (LDA) algorithm to perform topic modeling on Weibo texts from different regions, identifying the types of disasters affecting each township. Additionally, through correlation analysis, we investigate the relationship between disaster rescue requests and response measures to evaluate the adequacy of flood response measures in each township. The research results demonstrate that this analytical framework can accurately extract disaster information, precisely identify critical time points in flood disasters, locate core affected areas, uncover primary regional issues, and further validate the sufficiency of response measures, therefore enhancing the efficiency in collecting disaster information and analytical capabilities.

OCT monitoring data processing method of laser deep penetration welding based on HDBSCAN

To address the challenge of low accuracy in the measurement of laser deep penetration welding depth, this study introduces optical coherence tomography (OCT) technology for the direct measurement of this parameter, which is designed to enable real-time measurement and the accurate acquisition of depth information. This study has developed a laser welding depth inspection system that integrates Spectral-Domain OCT (SD-OCT), utilizing 304 stainless steel as the test material for inspecting laser deep penetration welding. A comparative analysis evaluated the impact of Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN), Density-Based Spatial Clustering of Applications with Noise (DBSCAN), and K-means clustering (K-mean) on the clustering of OCT raw data, in addition to the use of percentile filters combined with the exponential moving average method for extracting the depth-of-melt curve. The depth-of-melt curve, obtained via the HDBSCAN algorithm, more accurately reflects the actual depth curve, with an average error remaining below 5%. This error rate is 46% lower than that of DBSCAN and 42% lower than that of K-mean. The results of the experiments with continuously varying process parameters show that a high degree of accuracy can be demonstrated even when the welding process parameters are continuously varied. Experimental results indicate that the combination of the HDBSCAN algorithm with percentile filtering and the exponential moving average method significantly enhances the recognition accuracy of the deep penetration laser welding depth curve. This method provides more precise feature parameters for the real-time detection and identification of defects in deep penetration laser welding, thereby offering strong support for optimizing and controlling the laser welding process.

Membership Study of Open Cluster NGC 6134 using HDBSCAN Clustering Algorithm

Abstract We present the results of our study on the membership of open star clusters in the southern hemisphere, NGC 6134. The position of NGC 6134 is located near the galaxy’s plane, so the apparent position in the sky is in an area with high stellar density. This poses a challenge in distinguishing between cluster member stars and field stars. For this purpose, we use the Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm on astrometric data from Gaia Data Release 3 (DR3). We take stars as samples up to a distance of 90 parsecs from the cluster’s center as membership determination samples. The spatial scope of this sample is intentionally larger than previous studies on this star cluster, as some recent membership studies have concluded that the size of star clusters turns out to be larger than what has been reported. The results of our study produced a consistent number of members with the results of other studies with the same sample size. The clustering algorithm was then performed multiple times using different values of the radius sample. We determined that NGC 6134 has 888 members (membership probability ≥ 0.5), with a radius of 30 parsecs from the cluster’s center.

Predicting possible recommendations related to causes and consequences in the HAZOP study worksheet using natural language processing and machine learning: BERT, clustering, and classification

A set of recommendations is one of the most valuable outputs of the hazard and operability (HAZOP) study. The HAZOP study team provides recommendations when deficiencies are detected in the chemical process plant. These deficiencies can cause chemical process accidents and operability issues. This study employed a data-driven approach using natural language processing (NLP) and machine learning (ML) to predict potential recommendations based on causes and consequences. The dataset had no label; thus, clustering was used to label it. Firstly, bidirectional encoder representations from transformers (BERT) converted recommendation sentences into vectors. Secondly, uniform manifold approximation and projection (UMAP) and hierarchical density-based spatial clustering of applications with noise (HDBSCAN) were utilized to determine recommendation categories and label the dataset. Then, BERT was used to convert causes and consequences into vectors. Finally, a multi-layer perceptron (MLP) classifier was employed to predict possible recommendations based on causes and consequences. The class imbalance problem was handled by random over-sampling. The prediction accuracy of possible recommendations based on causes and consequences equals 93.7% and 89.5%, respectively. As a result of predicting potential recommendations utilizing causes and consequences, major recommendations will not be overlooked during the HAZOP study. This can further expand NLP and ML applications in HAZOP study automation.

A data-adaptive network design for the regional gravity field modelling using spherical radial basis functions

A high-precision regional gravity field model is significant in various geodesy applications. In the field of modelling regional gravity fields, the spherical radial basis functions (SRBFs) approach has recently gained widespread attention, while the modelling precision is primarily influenced by the base function network. In this study, we propose a method for constructing a data-adaptive network of SRBFs using a modified Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm, and the performance of the algorithm is verified by the observed gravity data in the Auvergne area. Furthermore, the turning point method is used to optimize the bandwidth of the basis function spectrum, which satisfies the demand for both high-precision gravity field and quasi-geoid modelling simultaneously. Numerical experimental results indicate that our algorithm has an accuracy of about 1.58 mGal in constructing the gravity field model and about 0.03 m in the regional quasi-geoid model. Compared to the existing methods, the number of SRBFs used for modelling has been reduced by 15.8%, and the time cost to determine the centre positions of SRBFs has been saved by 12.5%. Hence, the modified HDBSCAN algorithm presented here is a suitable design method for constructing the SRBF data adaptive network.

Modified HDBSCAN based segmentation hyperspectral image segmentation for cotton crop classification

ABSTRACT The most crucial element in accurately monitoring and assessing cotton development is having effective cotton maps. In order to make decisions about governance, precision agriculture, and field management, the county-scale cotton remote sensing categorisation models must be evaluated. The main objective of this research is to propose novel hyperspectral image segmentation approach for cotton crops to monitor the crops and identify early signs of disease. The proposal for a hyperspectral image-based classification of cotton crops is made in this research. Using ‘Modified Hierarchical density-based spatial clustering of applications with noise (HDBSCAN),’ the procedure begins with the input image being segmented. Following this, features based on vegetation indices, hybrid vegetation indices, and statistical characteristics will be retrieved and trained with the classification model to ensure proper classification. Specifically, EVI, NDVI, and RVI are features that are based on vegetation indices. Using techniques like SVM, CNN, DBN, DT, and Improved Bidirectional Long Short-Term Memory (IBi-LSTM), this study replicates a stacked ensemble framework for classification. While the MHDBSCAN achieved the maximum accuracy value of 97.97%, the conventional techniques achieved limited accuracy. Thus, the MHDBSCAN far more effective at classifying the crop utilising hyperspectral image segmentation and the classification become more precise and accurate.

Interrogation of Ecotoxic Elements Distribution in Slag and Precipitated Calcite through a Machine Learning‐Based Approach Aided by Mass Spectrometry

AbstractCO2 mineralization in slag has been widely investigated as a potential solution for offsetting steelmaking industry emissions. However, it can be associated with ecotoxic elements release (e.g., V and Cr). The presence of such elements in heterogenous slag at the micro‐scale remains difficult for analysis since microstructural features can be missed during microscopy data inspection, thereby presenting a challenge in understanding how ecotoxic elements exist in slag. Here, an unsupervised machine learning‐based technique is used to analyze slag's microstructural features. Energy Dispersive Spectroscopy (EDS) data are analyzed through Hierarchical Density‐Based Spatial Clustering of Applications with Noise (HDBSCAN) method. Results show that passive CO2 mineralization has occurred in situ in the studied samples, on the surface, and within their pores. Additionally, V and Cr regions with equivalent diameters < 42 µm can exist within slag, potentially making such elements prone to mobilization due to slag pulverization. Interrogation of the samples with Laser Ablation Inductively Coupled Plasma Mass Spectroscopy (LA‐ICP‐MS) confirms the distribution of the elements obtained from the clustering algorithm and further demonstrates that up to 84 and 9 ppm of V and Cr are incorporated in the precipitated calcite, respectively. This implies that ecotoxic elements may be immobilized through calcite precipitation.

Analysis of Urban Residents’ Travelling Characteristics and Hotspots Based on Taxi Trajectory Data

This study utilizes taxi trajectory data to uncover urban residents’ travel patterns, offering critical insights into the spatial and temporal dynamics of urban mobility. A fusion clustering algorithm is introduced, enhancing the clustering accuracy of trajectory data. This approach integrates the hierarchical density-based spatial clustering of applications with noise (HDBSCAN) algorithm, modified to incorporate time factors, with kernel density analysis. The fusion algorithm demonstrates a higher noise point detection rate (15.85%) compared with the DBSCAN algorithm alone (7.31%), thus significantly reducing noise impact in kernel density analysis. Spatial correlation analysis between hotspot areas and paths uncovers distinct travel behaviors: During morning and afternoon peak hours on weekdays, travel times (19–40 min) exceed those on weekends (16–35 min). Morning peak hours see higher taxi utilization in residential and transportation hubs, with schools and commercial and government areas as primary destinations. Conversely, afternoon peaks show a trend towards dining and entertainment zones from the abovementioned places. In the evening rush, residents enjoy a vibrant nightlife, and there are numerous locations for picking up and dropping off people. A chi-square test on weekday travel data yields a p-value of 0.023, indicating a significant correlation between the distribution of travel hotspots and paths.

Point cloud completion network for 3D shapes with morphologically diverse structures

Point cloud completion is a challenging task that involves predicting missing parts in incomplete 3D shapes. While existing strategies have shown effectiveness on point cloud datasets with regular shapes and continuous surfaces, they struggled to manage the morphologically diverse structures commonly encountered in real-world scenarios. This research proposed a new point cloud completion method, called SegCompletion, to derive complete 3D geometries from a partial shape with different structures and discontinuous surfaces. To achieve this, morphological segmentation was introduced before point cloud completion by deep hierarchical feature learning on point sets, and thus, the complex morphological structure was segmented into regular shapes and continuous surfaces. Additionally, each instance of a point cloud that belonged to the same type of feature could also be effectively identified using HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise). Furthermore, the multiscale generative network achieved sophisticated patching of missing point clouds under the same geometric feature based on feature points. To compensate for the variance in the mean distances between the centers of the patches and their closest neighbors, a simple yet effective uniform loss was utilized. A number of experiments on ShapeNet and Pheno4D datasets have shown the performance of SegCompletion on public datasets. Moreover, the contribution of SegCompletion to our dataset (Cotton3D) was discussed. The experimental results demonstrated that SegCompletion performed better than existing methods reported in the literature.

Traditional and GenAI Text Analysis of COVID-19 Pandemic Trends in Hospital Community Benefits IRS Documentation

The coronavirus disease 2019 (COVID-19) pandemic presented unique challenges to the U.S. healthcare system, particularly for nonprofit U.S. hospitals that are obligated to provide community benefits in exchange for federal tax exemptions. We sought to examine how hospitals initiated, modified, or disbanded community benefits programming in response to the COVID-19 pandemic. We used the free-response text in Part IV of Internal Revenue Service (IRS) Form 990 Schedule H (F990H) to assess health equity and disparities. We combined traditional key term frequency and Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) clustering approaches with a novel Generative Pre-trained Transformer (GPT) 3.5 summarization approach. Our research reveals shifts in community benefits programming. We observed an increase in COVID-related terms starting in the 2019 tax year, indicating a pivot in community focus and efforts toward pandemic-related activities such as telehealth services and COVID-19 testing and prevention. The clustering analysis identified themes related to COVID-19 and community benefits. Generative Artificial Intelligence (GenAI) summarization with GPT3.5 contextualized these changes, revealing examples of healthcare system adaptations and program cancellations. However, GPT3.5 also encountered some accuracy and validation challenges. This multifaceted text analysis underscores the adaptability of hospitals in maintaining community health support during crises and suggests the potential of advanced AI tools in evaluating large-scale qualitative data for policy and public health research.

Land use and sexual harassment: A geospatial analysis based on the volunteer HarassMap‐Egypt

AbstractSexual harassment and gang rape in Egypt have garnered attention from both traditional and digital media. This study employed a volunteer HarassMap to analyse sexual harassment crimes (SHCs) across Egypt from a spatial perspective. The specific aims were to apply the Hierarchical Density‐Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm to locate clusters of reported SHCs, and to assess their spatial dependence on land use types. To accomplish this task, ring buffers of 100, 200, 300, 400, and 500 metres were established around each crime scene to determine which land use was mostly associated with the incidence of these SHCs. Local bivariate relationships were used to explore the associations between SHC and each land‐use category. Results from the HDBSCAN algorithm revealed four crime clusters within the study domain, mainly located in Greater Cairo, Alexandria, and Behaira. Notably, commercial establishments and transit stations showed a significantly positive correlation with SHC. The study shows how land uses shape SHC and showed that it is possible to identify environmental risk factors for harassment. These risk factors can help policymakers, urban planners, and community stakeholders prevent and reduce sexual harassment and gender inequality, and promote just and inclusive societies.

Ship Trajectory Prediction: An Integrated Approach Using ConvLSTM-Based Sequence-to-Sequence Model

Maritime transportation is one of the major contributors to the development of the global economy. To ensure its safety and reduce the occurrence of a maritime accident, intelligent maritime monitoring and ship behavior identification have been drawing much attention from industry and academia, among which, the accurate prediction of ship trajectory is one of the key questions. This paper proposed a trajectory prediction model integrating the Convolutional LSTM (ConvLSTM) and Sequence to Sequence (Seq2Seq) models to facilitate simultaneous extraction of temporal and spatial features of ship trajectories, thereby enhancing the accuracy of prediction. Firstly, the trajectories are preprocessed using kinematic-based anomaly removal and Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) to improve the data quality for the training process of trajectory prediction. Secondly, the ConvLSTM-based Seq2seq model is designed to extract temporal and spatial features of the ship trajectory and improve the performance of long-time prediction. Finally, by using real AIS data, the proposed model is compared with the Seq2Seq and Bidirectional LSTM based on attention mechanism (Bi-Attention-LSTM) models to verify its effectiveness. The experimental results demonstrate that the proposed model achieves excellent performance in predicting turning trajectories, good predictive accuracy on straight line motions, and greater improvement in prediction accuracy compared to the other two benchmark models. Overall, the proposed model represents a promising contribution to improving ship trajectory prediction accuracy and may enhance the safety and quality of ship navigation in complex and volatile marine environments.

Unstable Approach Detection and Analysis Based on Energy Management and a Deep Neural Network

The study of managing risk in aviation is the key to improving flight safety. Compared to the other flight operation phases, the approach and landing phases are more critical and dangerous. This study aims to detect and analyze unstable approaches in Taiwan through historical flight data. In addition to weather factors such as low visibility and crosswinds, human factors also account for a large part of the risk. From the accidents studied in the stochastic report of the Flight Safety Foundation, nearly 70% of the accidents occurred during the approach and landing phases, which were caused by improper control of aircraft energy. Since the information of the flight data recorder (FDR) is regarded as the airline’s confidential information, this study calculates the aircraft’s energy-related metrics and investigates the influence of non-weather-related factors on unstable approaches through a publicly available source, automatic dependent surveillance-broadcast (ADS-B) flight data. To evaluate the influence of weather- and non-weather-related factors, the outliers of each group classified by weather labels are detected and eliminated from the analysis by applying hierarchical density-based spatial clustering of applications with noise (HDBSCAN), which is utilized for detecting abnormal flights that are spatial anomalies. The deep learning method was adopted to detect and predict unstable arrival flights landing at Taipei Songshan Airport. The accuracy of the prediction for the normalized total energy and trajectory deviation of all flights is 85.15% and 82.11%, respectively. The results show that in different kinds of weather conditions, or not considering the weather, the models have similar good performance. The input features were analyzed after the model was obtained, and the flights detected as abnormal are discussed.

Improving the open cluster census

Context.Data from theGaiasatellite are revolutionising our understanding of the Milky Way. With every new data release, there is a need to update the census of open clusters.Aims.We aim to conduct a blind, all-sky search for open clusters using 729 million sources fromGaiaDR3 down to magnitudeG ∼ 20, creating a homogeneous catalogue of clusters including many new objects.Methods.We used the Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm to recover clusters. We validated our clusters using a statistical density test and a Bayesian convolutional neural network for colour-magnitude diagram classification. We inferred basic astrometric parameters, ages, extinctions, and distances for the clusters in the catalogue.Results.We recovered 7167 clusters, 2387 of which are candidate new objects and 4782 of which crossmatch to objects in the literature, including 134 globular clusters. A more stringent cut of our catalogue contains 4105 highly reliable clusters, 739 of which are new. Owing to the scope of our methodology, we are able to tentatively suggest that many of the clusters we are unable to detect may not be real, including 1152 clusters from the Milky Way Star Cluster (MWSC) catalogue that should have been detectable inGaiadata. Our cluster membership lists include many new members and often include tidal tails. Our catalogue’s distribution traces the galactic warp, the spiral arm structure, and the dust distribution of the Milky Way. While much of the content of our catalogue contains bound open and globular clusters, as many as a few thousand of our clusters are more compatible with unbound moving groups, which we will classify in an upcoming work.Conclusions.We have conducted the largest search for open clusters to date, producing a single homogeneous star cluster catalogue which we make available with this paper.

Predicting Measles Outbreaks in the United States: Evaluation of Machine Learning Approaches.

Measles, a highly contagious viral infection, is resurging in the United States, driven by international importation and declining domestic vaccination coverage. Despite this resurgence, measles outbreaks are still rare events that are difficult to predict. Improved methods to predict outbreaks at the county level would facilitate the optimal allocation of public health resources. We aimed to validate and compare extreme gradient boosting (XGBoost) and logistic regression, 2 supervised learning approaches, to predict the US counties most likely to experience measles cases. We also aimed to assess the performance of hybrid versions of these models that incorporated additional predictors generated by 2 clustering algorithms, hierarchical density-based spatial clustering of applications with noise (HDBSCAN) and unsupervised random forest (uRF). We constructed a supervised machine learning model based on XGBoost and unsupervised models based on HDBSCAN and uRF. The unsupervised models were used to investigate clustering patterns among counties with measles outbreaks; these clustering data were also incorporated into hybrid XGBoost models as additional input variables. The machine learning models were then compared to logistic regression models with and without input from the unsupervised models. Both HDBSCAN and uRF identified clusters that included a high percentage of counties with measles outbreaks. XGBoost and XGBoost hybrid models outperformed logistic regression and logistic regression hybrid models, with the area under the receiver operating curve values of 0.920-0.926 versus 0.900-0.908, the area under the precision-recall curve values of 0.522-0.532 versus 0.485-0.513, and F2 scores of 0.595-0.601 versus 0.385-0.426. Logistic regression or logistic regression hybrid models had higher sensitivity than XGBoost or XGBoost hybrid models (0.837-0.857 vs 0.704-0.735) but a lower positive predictive value (0.122-0.141 vs 0.340-0.367) and specificity (0.793-0.821 vs 0.952-0.958). The hybrid versions of the logistic regression and XGBoost models had slightly higher areas under the precision-recall curve, specificity, and positive predictive values than the respective models that did not include any unsupervised features. XGBoost provided more accurate predictions of measles cases at the county level compared with logistic regression. The threshold of prediction in this model can be adjusted to align with each county's resources, priorities, and risk for measles. While clustering pattern data from unsupervised machine learning approaches improved some aspects of model performance in this imbalanced data set, the optimal approach for the integration of such approaches with supervised machine learning models requires further investigation.

Automatic individual recognition of wild Crested Ibis based on hybrid method of self-supervised learning and clustering

Automatic individual recognition of wild Crested Ibis is challenged by two factors: the lack of labeled data and an unpredictable number of individuals. In this paper, we propose a hybrid method of self-supervised learning and clustering to automatically recognize wild Crested Ibis based on vocalizations. To address the first challenge, we enhance the Bootstrap Your Own Latent for Audio (BYOL-A) model by using an improved augmentation module and Spatial Group-wise Enhance (SGE) attention module to create the self-supervised learning model BYOL-AIS. This model aims to extract a more discriminative representation of Crested Ibis vocalizations. To handle the second challenge, we introduce a clustering method that combines Uniform Manifold Approximation and Projection (UMAP) and Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) to distinguish the representation of Crested Ibis vocalizations. We evaluate our proposed method using vocalizations collected from 10 Crested Ibis individuals and achieve a recognition accuracy of 0.864. This accuracy is comparable to the performance of commonly used supervised methods. Our results suggest that our proposed method is a feasible method for wild bird recognition in the absence of labeled data and has the potential to be an analytical tool for processing huge amounts of monitoring data.