Unsupervised Classification Method Research Articles

SPATIO-TEMPORAL DYNAMICS OF URBAN THERMAL ENVIRONMENT IN UDAIPUR CITY, RAJASTHAN, INDIA

Abstract. The Udaipur urban agglomeration was selected to analyse the urban heat island (UHI) effect in the area from 2005 to 2017. Landsat 7 ETM+ sensor data was used to derive land surface temperatures (LST) and map landscape characteristics. The agglomeration was classified into seven land use land cover classes including agriculture cropped, agriculture fallow, barren, built up, scrub, vegetation and water using unsupervised classification method. The LST results were obtained using NDVI method. Urban heat island intensities were mapped using z-score method. Regions with UHI values of above 2 standard deviations were considered to reflect UHI effect. Results show an overall decrease of 10.2 percent in agricultural cropped and fallow lands. The scrub class also shows a moderate decline of 2.3 percent. Increase in area were observed for built up, barren and vegetation classes by 8.8, 2.1 and 1.9 percent respectively. The dominant LULC change was the transformation of agricultural lands to built-up class. The class-wise mean LST observations increase in order of water, vegetation, cropped, fallow, built up, scrub and barren land. The mean LST records an increase of 1.7° C from 30.1° C to 31.8° C over the study period. Results suggest that UHI effects are more prominent in urban fringe area corresponding with the built up and barren land cover classes contrary to its core area. Built up areas surrounded by barren lands left vacant for city’s sprawl in the future exhibit highest UHI intensities.

Unsupervised domain adaptation for SAR target classification based on domain- and class-level alignment: From simulated to real data

Synthetic Aperture Radar (SAR) target classification methods based on convolutional neural networks have attracted wide attention recently. Typically, these methods necessitate a substantial quantity of labeled data to train the deep model, which is an obstacle for SAR target classification since annotating SAR images is costly and time-consuming. Generating large-scale labeled data in simulation to train the deep model is a feasible alternative. However, cross-domain recognition is challenging due to significant differences in distribution between simulated and real domains. In this paper, we propose an unsupervised domain adaptation method for SAR target classification by aligning the distribution between the simulated and unlabeled real domains. Specifically, we encode the dependencies across different granularity perspectives including domain-, and class-levels simultaneously to align two domains finely. On the one hand, a gradient-weighted adversarial alignment method is proposed to align the global distributions at the domain-level. On the other hand, a novel class-level alignment method based on the prototypical network is proposed to align the finer class structure. Technically, we respectively model the cross-domain prototype-prototype relations and cross-domain prototype-instance relations with contrastive learning to achieve intra-class prototype compact as well as inter-class prototype separation among cross-domain categories, where the prototypes of the real domain are learned with pseudo labels. To improve the quality of pseudo-labels, a pseudo-label filtering strategy combined with the attribute scattering center is further designed. The experiments on two simulation-to-reality datasets indicate that the proposed method outperforms state-of-the-art unsupervised domain adaptive methods, such as pixelfeat (+4.45% and +7.6% in accuracy on SAMPLE and S2M datasets respectively). Code and data will be available in https://github.com/YuShi1213/Sim2Real-Unsupervised-SAR-Target-Classification.

Unsupervised Classification under Uncertainty: The Distance-Based Algorithm

This paper presents a method for unsupervised classification of entities by a group of agents with unknown domains and levels of expertise. In contrast to the existing methods based on majority voting (“wisdom of the crowd”) and their extensions by expectation-maximization procedures, the suggested method first determines the levels of the agents’ expertise and then weights their opinions by their expertise level. In particular, we assume that agents will have relatively closer classifications in their field of expertise. Therefore, the expert agents are recognized by using a weighted Hamming distance between their classifications, and then the final classification of the group is determined from the agents’ classifications by expectation-maximization techniques, with preference to the recognized experts. The algorithm was verified and tested on simulated and real-world datasets and benchmarked against known existing algorithms. We show that such a method reduces incorrect classifications and effectively solves the problem of unsupervised collaborative classification under uncertainty, while outperforming other known methods.

Gold mining's environmental footprints, drivers, and future predictions in Ghana

The last two decades have seen a surge in gold mining operations around the world. Despite mining occupying a smaller geographical area compared to other land use/land cover (LULC) classes, it exhibits strong interconnections with various land uses and serves as a major driver for changes in mining landscapes. Understanding and evaluating historical and potential future LULC changes in these landscapes are crucial in assessing the environmental impact of mining. Traditionally, these assessments heavily rely on geospatial techniques, with limited emphasis on projecting future LULC trends. This research aims to monitor, analyse the drivers of change, and predict future changes in LULC under two scenarios: the “business as usual” scenario and the "remedial measures" scenarios. Utilising the CA-Markov model, this article predicts LULC changes and offers comprehensive insights into the environmental impacts of mining, combining geospatial and social research methodologies. The investigation spanned a 34-year period (1986–2020) and employed a blend of supervised and unsupervised image classification methods, complemented by interviews, focus groups, and field observations. The findings reveal substantial land degradation, water pollution, and a significant loss of forest cover, accounting for 27,333 ha (36%). Continuation of current mining practices is predicted to lead to further ecological deterioration.

A Novel Fuzzy Unsupervised Quadratic Surface Support Vector Machine Based on DC Programming: An Application to Credit Risk Management

Unsupervised classification is used in credit risk assessment to reduce human resource costs and make informed decisions in the shortest possible time. Although several studies show that support vector machine-based methods have better performance in unlabeled datasets, several factors still negatively affect these models, such as unstable results due to random initialization, reduced effectiveness due to kernel dependencies, and noise points and outliers. This paper introduces an unsupervised classification method based on a fuzzy unsupervised quadratic surface support vector machine without a kernel to avoid selecting related kernel parameters for credit risk assessment. In addition, we propose an innovative fuzzy membership function for reducing noise points and outliers in line with the direction of sample density variation. Fuzzy Unsupervised QSSVM (FUS-QSSVM) outperforms well-known SVM-based methods based on numerical tests on public benchmark credit data. In some real-world applications, the proposed method has significant potential as well as being effective, efficient, and robust. The algorithm can therefore increase the number of potential customers of financial institutions as well as increase profitability.

Machine Learning Approach for Catastrophe Risk Assessment and Management Using Remote Sensing Data

With emergency programs for disaster preparedness and warning phases for earthquakes, landslides, and floods in recent years, remote sensing and geographic information systems have played a crucial role in Catastrophe Risk assessment and management. It has also been a key focus in the field of technology. Without the right tool for organizing massive volumes of data and gathering information from many sources, such maps or measurement channels, it would not be feasible to employ sensory data. In order to identify and assess areas affected by floods, earthquakes, avalanches, landslides, and wildfires, this study employs machine learning approaches. Following the application of filters to enhance image quality, the images are segmented through thresholding technique and classified using supervised and unsupervised classification methods. Images from before and after disasters are gathered from MRSAC Nagpur and processed using Python-based tools, ArcGIS, ERDAS, and QGIS for the purpose of analyzing devastation.

Machine learning and data analysis for word segmentation of classical Chinese poems: illustrations with Tang and Song examples

Abstract Words are essential parts for understanding classical Chinese poems. We report a collection of 32,399 classical Chinese poems that were annotated with word boundaries. Statistics about the annotated poems support a few heuristic experiences, including the patterns of lines and a practice for the parallel structures (對仗), that researchers of Chinese literature discuss in the literature. The annotators were affiliated with two universities, so they could annotate the poems as independently as possible. Results of an inter-rater agreement study indicate that the annotators have consensus over the identified words 93 per cent of the time and have perfect consensus for the segmentation of a poem 42 per cent of the time. We applied unsupervised classification methods to annotate the poems in several different settings, and evaluated the results with human annotations. Under favorable conditions, the classifier identified about 88 per cent of the words, and segmented poems perfectly 22 per cent of the time.

Dynamic Graph Clustering Learning for Unsupervised Diabetic Retinopathy Classification.

Diabetic retinopathy (DR) is a common complication of diabetes, which can lead to vision loss. Early diagnosis is crucial to prevent the progression of DR. In recent years, deep learning approaches have shown promising results in the development of an intelligent and efficient system for DR classification. However, one major drawback is the need for expert-annotated datasets, which are both time-consuming and costly. To address these challenges, this paper proposes a novel dynamic graph clustering learning (DGCL) method for unsupervised classification of DR, which innovatively deploys the Euclidean and topological features from fundus images for dynamic clustering. Firstly, a multi-structural feature fusion (MFF) module extracts features from the structure of the fundus image and captures topological relationships among multiple samples, generating a fused representation. Secondly, another consistency smoothing clustering (CSC) module combines network updates and deep clustering to ensure stability and smooth performance improvement during model convergence, optimizing the clustering process by iteratively updating the network and refining the clustering results. Lastly, dynamic memory storage is utilized to track and store important information from previous iterations, enhancing the training stability and convergence. During validation, the experimental results with public datasets demonstrated the superiority of our proposed DGCL network.

The verification of hen egg types by the classification of ultra-weak photon emission data

Examples from living systems at various levels of the biological hierarchy and also from natural food products show that ultra-weak photon emission (UPE) has potential applications in the rating of vital functions and quality testing. In this study, the UPE of chicken eggs has been tested regarding the possibility of egg quality verification. The UPE from intact eggs and separated egg parts were subjected to supervised and unsupervised classification methods according to different housing types. The results of unsupervised egg grouping substantially agreed with the types of hen rearing. The Cohen’s Kappa test score for the K-means method was up to K=0.63. Supervised Support Vector Machine (SVM) classifier with radial kernel function achieved a relatively high accuracy (AC), up to 88%, also confirmed by the value of the K-statistics up to 0.81. This study shows that the best result of egg types classification can be obtained using UPE emission data from all egg parts.

Groundwater Availability and Suitability Evaluation for Irrigated Agriculture in Mbale District

The purpose of this study was to evaluate the availability and suitability of groundwater in Mbale district basing on potential, quality and vulnerability to support irrigated agriculture. Groundwater potential was assessed using geospatial technique through AHP model. Thematic layers were prepared namely: land use and land cover, slope, soil, rainfall, lineament density, lithology, geomorphology and drainage density. All layers were integrated using the MLC technique. Ranking of each parameter was performed using WOIA. Weights were assigned to each subject class basing on AHP results. The consistency of the outputs was tested by computation of CR and was at a reasonable acceptable level (0.029 < 0.1). Groundwater potential was delineated basing on the values of groundwater potential index. Groundwater quality was determined using IWQI method. Thirty-three water samples were collected. Five chemical parameters were tested in the laboratory: chloride, electrical conductivity, bicarbonate, sodium and SAR to generate the water quality database using Kriging interpolation technique. Computation of IWQI values for each source was made and was used to generate the IWQI map using the weighted summation. DRATIC method was used to delineate vulnerability of groundwater. Layers were generated using the following parameters: depth to groundwater, net recharge, aquifer media, topography, vadose zone impacts, and hydraulic conductivity. The aquifer vulnerability map was prepared by overlaying layers. Three different vulnerability zones were determined according to DRASTIC scores low (<100), medium (100-140) and high (>140). Integration of maps of quality, vulnerability and potential of groundwater was made using an unsupervised MLC classification method. Groundwater potential was in three zones: very good zone was 26.99 km2 (12.98%), the good zone was 126.22 km2 (60.71%) and the poor zone was 54.69 km2 (26.31%). The average annual exploitable groundwater reserves are estimated at 0.026 (MCM/km2) in the zone of “very good”, 0.024 MCM/km2 in the zone of “good” and 0.018 MCM/km2 in the zone of “poor”. Groundwater quality was in two use restrictions: High Restriction of 90.90% with the IWQI value from 40 to 55 on the area of 188.98 km2 of the study area and Severe Restriction of 9.10% with IWQI values from 32 to 40 covering of 18.92 km2. The study area had a high probability of contamination. Low vulnerability covers an area of 11.23km2 (5.40%), 77.83km2 (37.44%) for medium and 118.83km2 (57.16%) for high. The resulting clustered map was classified into five categories with their respective regions: 17.58% very poor (36.56km2), 13.84% poor (28.77km2), 12.69% good (26.39km2), 31.46% very good (65.39km2) and 24.43 % excellent (50.78km2). The sub counties in Mbale district that have inadequate and unsuitable groundwater for irrigated agriculture are Budwale, Wanale, Busano, Bubyangu, Bufumbo and Nyondo since they belong to the zones of very poor and poor. The remaining sub counties have available and suitable groundwater for irrigated agriculture since they belong to the zones of good, very good and excellent.

Automatic Core-Developer Identification on GitHub: A Validation Study

Many open-source software projects are self-organized and do not maintain official lists with information on developer roles. So, knowing which developers take core and maintainer roles is, despite being relevant, often tacit knowledge. We propose a method to automatically identify core developers based on role permissions of privileged events triggered in GitHub issues and pull requests. In an empirical study on 25/GitHub projects, (1) we validate the set of automatically identified core developers with a sample of project-reported developer lists, and (2) we use our set of identified core developers to assess the accuracy of state-of-the-art unsupervised developer classification methods. Our results indicate that the set of core developers, which we extracted from privileged issue events, is sound and the accuracy of state-of-the-art unsupervised classification methods depends mainly on the data source (commit data versus issue data) rather than the network-construction method (directed versus undirected, etc.). In perspective, our results shall guide research and practice to choose appropriate unsupervised classification methods, and our method can help create reliable ground-truth data for training supervised classification methods.

Improving Neutron-gamma discrimination in CLYC with Gaussian mixture model

Neutron and gamma discrimination is a common issue in neutron detection. Cs 2 LiYCl 6:Ce (CLYC) crystal, with its unique core-to-valence luminescence (CVL) mechanism for gamma rays, has excellent capability for neutron-gamma discrimination. This paper proposes a new preprocessing method called constant fraction alignment (CFA), based on the fast rise and slow decay characteristics of CLYC, to effectively improve discrimination performance. Unsupervised classification methods that do not require training data have a great advantage since pure neutron sources are difficult to obtain. As an unsupervised clustering method, Gaussian mixture model (GMM) is used to achieve neutron-gamma discrimination of data detected by CLYC, and the impact of different parameters on GMM results is studied. When the input dimension is high, GMM not only requires significantly more time but also reduces precision, so dimensionality reduction is necessary. Principal component analysis (PCA) is a commonly used dimensionality reduction method, and through analysis, it is found that using the first three components of PCA as inputs achieves the best performance. In the GMM solving process, selecting different covariance matrices based on data types can reduce running time while maintaining high accuracy. Diagonal mode is the most suitable according to comparison. Based on the same dataset, the accuracy of GMM is higher than that of charge comparison method and k-means++, with an accuracy of 99.96% for 252 Cf source and a false alarm rate as low as 0.01% for 137 Cs source.

Intelligent Unsupervised Network Traffic Classification Method Using Adversarial Training and Deep Clustering for Secure Internet of Things

Network traffic classification (NTC) has attracted great attention in many applications such as secure communications, intrusion detection systems. The existing NTC methods based on supervised learning rely on sufficient labeled datasets in the training phase, but for most traffic datasets, it is difficult to obtain label information in practical applications. Although unsupervised learning does not rely on labels, its classification accuracy is not high, and the number of data classes is difficult to determine. This paper proposes an unsupervised NTC method based on adversarial training and deep clustering with improved network traffic classification (NTC) and lower computational complexity in comparison with the traditional clustering algorithms. Here, the training process does not require data labels, which greatly reduce the computational complexity of the network traffic classification through pretraining. In the pretraining stage, an autoencoder (AE) is used to reduce the dimension of features and reduce the complexity of the initial high-dimensional network traffic data features. Moreover, we employ the adversarial training model and a deep clustering structure to further optimize the extracted features. The experimental results show that our proposed method has robust performance, with a multiclassification accuracy of 92.2%, which is suitable for classification with a large number of unlabeled data in actual application scenarios. This paper only focuses on breakthroughs in the algorithm stage, and future work can be focused on the deployment and adaptation in practical environments.

Generation and propagation of bursts of activity in the developing basal ganglia.

The neonatal brain is characterized by intermittent bursts of oscillatory activity interspersed by relative silence. Although well-characterized for many cortical areas, to what extent these propagate and interact with subcortical brain areas is largely unknown. Here, early network activity was recorded from the developing basal ganglia, including motor/somatosensory cortex, dorsal striatum, and intralaminar thalamus, during the first postnatal weeks in mice. An unsupervised detection and classification method revealed two main classes of bursting activity, namely spindle bursts and nested gamma spindle bursts, characterized by oscillatory activity at ~ 10 and ~ 30Hz frequencies, respectively. These were reliably identified across all three brain regions and exhibited region-specific differences in their structural, spectral, and developmental characteristics. Bursts of the same type often co-occurred in different brain regions and coherence and cross-correlation analyses reveal dynamic developmental changes in their interactions. The strongest interactions were seen for cortex and striatum, from the first postnatal week onwards, and cortex appeared to drive burst events in subcortical regions. Together, these results provide the first detailed description of early network activity within the developing basal ganglia and suggest that cortex is one of the main drivers of activity in downstream nuclei during this postnatal period.

The Effect of the 2004 EU Enlargement on the Development and Similarity of the Insurance Sectors in the EU

The enlargement of the European Union to new countries in 2004 introduced mechanisms to support the development of various social and economic areas, while also aiming to level the differences between the member states. The primary purpose of this study is to analyse the development and similarities of the insurance markets in the old and new member states of the EU after the enlargement in 2004. We examine the insurance sector of both groups of countries, i.e., those that were members before 2004 and those that joined in 2004, using Hellwig’s development measure, which takes into account several characteristics. Additionally, we analyse the similarity of these countries using three statistical methods of unsupervised classification: Ward’s method, the k-means method, and the Partitioning Among Medoids. Our results indicate that there was a significant variation in the insurance characteristics of these compared groups of EU members after 2004. In general, the insurance markets of the old and new EU countries developed differently. Since the enlargement in 2004, the insurance markets of the old and new EU countries have not yet aligned.

FedUSC: Collaborative Unsupervised Representation Learning From Decentralized Data for Internet of Things

Federated Learning (FL) lately has shown much promise in improving the shared model and preserving data privacy. However, these existing methods are only of limited utility in the Internet of Things (IoT) scenarios, as they either heavily depend on high-quality labeled data or only perform well under idealized conditions, which typically cannot be found in practical applications. In this paper, we propose a novel federated unsupervised learning method for image classification without the use of any ground truth annotations. In IoT scenarios, a big challenge is that decentralized data among multiple clients is normally non-IID, leading to performance degradation. To address this issue, we further propose a dynamic update mechanism that can decide how to update the local model based on weights divergence. Extensive experiments show that our method outperforms all baseline methods by large margins, including +6.67% on CIFAR-10, +5.15% on STL-10, and +8.44% on SVHN in terms of classification accuracy. In particular, we obtain promising results on Mini-ImageNet and COVID-19 datasets and outperform several federated unsupervised learning methods under non-IID settings.

Urban Expansion and The Microclimate Characteristics Evolution of Lhasa

Urban expansion is the process of growing urban population and expanding land for residential and productive use. Microclimates are small-scale climates with relatively uniform and changeable climatic characteristics. In the context of climate change on the Tibetan Plateau, the microclimate of Lhasa will respond to urban expansion. With the support of landset8 OLI_TIRS data, this study uses the unsupervised classification method ISO-data algorithm to extract information on construction land and calculate the intensity of urban sprawl. In order to analyze and investigate the evolution of urban expansion and microclimate characteristics in Lhasa city from 2014 to 2021, the radiative transfer equation method was chosen to invert the surface temperature and calculate the heat island effect characteristic-related index. The results of the study show that the area of built-up land in Lhasa city increases and the heat island intensity rises during 2014-2021. This study contributes to urban planning in Lhasa and provides a case study of climate change and cities on the Tibetan Plateau.

The Effectiveness of Remote Sensing Techniques for Land use Classification in Kota Belud, Sabah

The objective of this study was to evaluate the effectiveness of remote sensing techniques in classifying land use types in Kota Belud, Sabah. Two classification techniques, unsupervised and supervised, were employed to ensure high accuracy of the land use data. Six land use types were classified, including paddy fields, bare land, water bodies, forest, urban areas, and oil palm plantations. This study found that the accuracy of land use classification for the years 1990, 2000, 2010, and 2020 was 94%, 86%, 98.30%, and 91.60%, respectively. These results suggest that the supervised classification method was more effective in accurately classifying land use types compared to the unsupervised classification method. The final maps generated by both classification methods showed significant changes in land use over time. For instance, the expansion of paddy fields was observed in the study area, particularly from 1990 to 2020. The conversion of forest areas and bare land to paddy cultivation was in line with the National Agro-Food Policy 2021-2030, which aims to make Kota Belud a paddy granary. Overall, the study highlights the importance of remote sensing techniques in monitoring land use changes. The results provide valuable insights for policymakers and land use planners in making informed decisions to manage land resources sustainably.

Indications of marine benthos occurrence from multi-spectral multi-beam backscatter data: a case study in the North Sea

To facilitate the conservation of seafloor habitats and planning of offshore activities, there is a growing need for mapping marine benthos in an effective and efficient way. Acoustic data acquired by multi-beam echosounders (MBES) have been extensively used for large-scale and high-resolution seafloor characterization. A deeper understanding of the relationship between backscatter data and sediment compositions can help to identify the benthos occurrence from the MBES data. With two multi-spectral MBES datasets collected near the western Wadden Sea islands in the North Sea, we investigated the potential of mapping marine benthos through backscatter classification. Two unsupervised classification methods, i.e., Bayesian classification, which mainly exploits the backscatter strength from incident angles larger than 20°, and hierarchical clustering of the backscatter strength at different angular ranges, were employed and the results were compared. The classification results from both methods showed a good correspondence with sediment properties such as the median grain size. Moreover, based on a principal component analysis of bottom sample properties, the hierarchical clustering results indicated a better distinction between contributions from the gravel content and benthos occurrence, e.g., sand mason worm density, than Bayesian classification, through involving the backscatter angular variations. Classification for multiple frequencies, on the other hand, showed little difference regarding the relationship with bottom sample properties. Although the backscatter difference between frequencies was also found to positively correlate with certain sample properties, using multi-spectral features for acoustic classification in this study did not reveal additional information compared to single-frequency classification results.

Identifying High-Need Primary Care Patients Using Nursing Knowledge and Machine Learning Methods.

Patient cohorts generated by machine learning can be enhanced with clinical knowledge to increase translational value and provide a practical approach to patient segmentation based on a mix of medical, behavioral, and social factors. This study aimed to generate a pragmatic example of how machine learning could be used to quickly and meaningfully cohort patients using unsupervised classification methods. Additionally, to demonstrate increased translational value of machine learning models through the integration of nursing knowledge. A primary care practice dataset (N = 3,438) of high-need patients defined by practice criteria was parsed to a subset population of patients with diabetes (n = 1233). Three expert nurses selected variables for k-means cluster analysis using knowledge of critical factors for care coordination. Nursing knowledge was again applied to describe the psychosocial phenotypes in four prominent clusters, aligned with social and medical care plans. Four distinct clusters interpreted and mapped to psychosocial need profiles, allowing for immediate translation to clinical practice through the creation of actionable social and medical care plans. (1) A large cluster of racially diverse female, non-English speakers with low medical complexity, and history of childhood illness; (2) a large cluster of English speakers with significant comorbidities (obesity and respiratory disease); (3) a small cluster of males with substance use disorder and significant comorbidities (mental health, liver and cardiovascular disease) who frequently visit the hospital; and (4) a moderate cluster of older, racially diverse patients with renal failure. This manuscript provides a practical method for analysis of primary care practice data using machine learning in tandem with expert clinical knowledge.