Unsupervised Learning Clustering Research Articles

Deep Adaptive Fuzzy Clustering for Evolutionary Unsupervised Representation Learning.

Cluster assignment of large and complex datasets is a crucial but challenging task in pattern recognition and computer vision. In this study, we explore the possibility of employing fuzzy clustering in a deep neural network framework. Thus, we present a novel evolutionary unsupervised learning representation model with iterative optimization. It implements the deep adaptive fuzzy clustering (DAFC) strategy that learns a convolutional neural network classifier from given only unlabeled data samples. DAFC consists of a deep feature quality-verifying model and a fuzzy clustering model, where deep feature representation learning loss function and embedded fuzzy clustering with the weighted adaptive entropy is implemented. We joint fuzzy clustering to the deep reconstruction model, in which fuzzy membership is utilized to represent a clear structure of deep cluster assignments and jointly optimize for the deep representation learning and clustering. Also, the joint model evaluates current clustering performance by inspecting whether the resampled data from estimated bottleneck space have consistent clustering properties to improve the deep clustering model progressively. Experiments on various datasets show that the proposed method obtains a substantially better performance for both reconstruction and clustering quality compared to the other state-of-the-art deep clustering methods, as demonstrated with the in-depth analysis in the extensive experiments.

Wave energy resource classification system for the China East Adjacent Seas based on multivariate clustering

This study based on the 25-year wave hindcast database of the western Pacific and used three unsupervised learning clustering algorithms to classify the wave energy resources in the China East Adjacent Seas. Five wave energy characteristic parameters are comprehensively considered in the calculation process of the clustering algorithm. According to the analysis of the classification results, it can be seen that the Class Ⅳ is the most suitable for wave energy development in the China East Adjacent Seas, followed by the Class Ⅲ and Class Ⅴ. The Class Ⅵ is too far away from the coast to be used as the intended area for wave energy development. The Class Ⅰ is mostly located in inland seas and harbors, which are not suitable for wave energy development. By analyzing the annual average captured power, it can be seen that the optimal capture interval of the existing wave energy converters with mature technology is too large compared with the wave conditions in the China East Adjacent Seas. We should vigorously develop wave energy converters that are more suitable for the wave conditions of Class Ⅲ, Class Ⅳ and Class Ⅴ, and improve the capture efficiency of the wave energy converters.

Using unsupervised learning to classify inlet water for more stable design of water reuse in industrial parks.

The water reuse facilities of industrial parks face the challenge of managing a growing variety of wastewater sources as their inlet water. Typically, this clustering outcome is designed by engineers with extensive expertise. This paper presents an innovative application of unsupervised learning methods to classify inlet water in Chinese water reuse stations, aiming to reduce reliance on engineer experience. The concept of 'water quality distance' was incorporated into three unsupervised learning clustering algorithms (K-means, DBSCAN, and AGNES), which were validated through six case studies. Of the six cases, three were employed to illustrate the feasibility of the unsupervised learning clustering algorithm. The results indicated that the clustering algorithm exhibited greater stability and excellence compared to both artificial clustering and ChatGPT-based clustering. The remaining three cases were utilized to showcase the reliability of the three clustering algorithms. The findings revealed that the AGNES algorithm demonstrated superior potential application ability. The average purity in six cases of K-means, DBSCAN, and AGNES were 0.947, 0.852, and 0.955, respectively.

Spatiotemporal evolution of COVID-19 in Portugal’s Mainland with self-organizing maps

BackgroundSelf-Organizing Maps (SOM) are an unsupervised learning clustering and dimensionality reduction algorithm capable of mapping an initial complex high-dimensional data set into a low-dimensional domain, such as a two-dimensional grid of neurons. In the reduced space, the original complex patterns and their interactions can be better visualized, interpreted and understood.MethodsWe use SOM to simultaneously couple the spatial and temporal domains of the COVID-19 evolution in the 278 municipalities of mainland Portugal during the first year of the pandemic. Temporal 14-days cumulative incidence time series along with socio-economic and demographic indicators per municipality were analyzed with SOM to identify regions of the country with similar behavior and infer the possible common origins of the incidence evolution.ResultsThe results show how neighbor municipalities tend to share a similar behavior of the disease, revealing the strong spatiotemporal relationship of the COVID-19 spreading beyond the administrative borders of each municipality. Additionally, we demonstrate how local socio-economic and demographic characteristics evolved as determinants of COVID-19 transmission, during the 1st wave school density per municipality was more relevant, where during 2nd wave jobs in the secondary sector and the deprivation score were more relevant.ConclusionsThe results show that SOM can be an effective tool to analysing the spatiotemporal behavior of COVID-19 and synthetize the history of the disease in mainland Portugal during the period in analysis. While SOM have been applied to diverse scientific fields, the application of SOM to study the spatiotemporal evolution of COVID-19 is still limited. This work illustrates how SOM can be used to describe the spatiotemporal behavior of epidemic events. While the example shown herein uses 14-days cumulative incidence curves, the same analysis can be performed using other relevant data such as mortality data, vaccination rates or even infection rates of other disease of infectious nature.

A machine learning-based predictive approach in evaluating consumer behavior

Organizations today want to be extrapolative; they want to gain information and insights on every minutia of customers. Predictive analytics is one of the tools which has proved to be a cornerstone in customer centricity. Though analytics is used for performing customer analysis for decades, the manual approach to data management and analysis has constrained the functionalities. Professionals now leverage the individual information from demographics to purchase history and perform predictive analytics for optimal decision-making, business marketing and thus business growth. Here we apply the ensemble learning method for classification based on the monthly income and spending behaviour of customers. Secondly, we analyse customer satisfaction by applying an unsupervised learning clustering method based on similarity through different performance measures.

Networked Microgrid Energy Management Based on Supervised and Unsupervised Learning Clustering

Networked microgrid (NMG) is a novel conceptual paradigm that can bring multiple advantages to the distributed system. Increasing renewable energy utilization, reliability and efficiency of system operation and flexibility of energy sharing amongst several microgrids (MGs) are some specific privileges of NMG. In this paper, residential MGs, commercial MGs, and industrial MGs are considered as a community of NMG. The loads’ profiles are split into multiple sections to evaluate the maximum load demand (MLD). Based on the optimal operation of each MG, the operating reserve (OR) of the MGs is calculated for each section. Then, the self-organizing map as a supervised and a k-means algorithm as an unsupervised learning clustering method is utilized to cluster the MGs and effective energy-sharing. The clustering is based on the maximum load demand of MGs and the operating reserve of dispatchable energy sources, and the goal is to provide a more efficient system with high reliability. Eventually, the performance of this energy management and its benefits to the whole system is surveyed effectively. The proposed energy management system offers a more reliable system due to the possibility of reserved energy for MGs in case of power outage variation or shortage of power.

Improvements for Therapeutic Intervention from the Use of Web Applications and Machine Learning Techniques in Different Affectations in Children Aged 0-6 Years.

Technological advances together with machine learning techniques give health science disciplines tools that can improve the accuracy of evaluation and diagnosis. The objectives of this study were: (1) to design a web application based on cloud technology (eEarlyCare-T) for creating personalized therapeutic intervention programs for children aged 0–6 years old; (2) to carry out a pilot study to test the usability of the eEarlyCare-T application in therapeutic intervention programs. We performed a pilot study with 23 children aged between 3 and 6 years old who presented a variety of developmental problems. In the data analysis, we used machine learning techniques of supervised learning (prediction) and unsupervised learning (clustering). Three clusters were found in terms of functional development in the 11 areas of development. Based on these groupings, various personalized therapeutic intervention plans were designed. The variable with most predictive value for functional development was the users’ developmental age (predicted 75% of the development in the various areas). The use of web applications together with machine learning techniques facilitates the analysis of functional development in young children and the proposal of personalized intervention programs.

Using an Unsupervised Clustering Model to Detect the Early Spread of SARS-CoV-2 Worldwide.

Deciphering the population structure of SARS-CoV-2 is critical to inform public health management and reduce the risk of future dissemination. With the continuous accruing of SARS-CoV-2 genomes worldwide, discovering an effective way to group these genomes is critical for organizing the landscape of the population structure of the virus. Taking advantage of recently published state-of-the-art machine learning algorithms, we used an unsupervised deep learning clustering algorithm to group a total of 16,873 SARS-CoV-2 genomes. Using single nucleotide polymorphisms as input features, we identified six major subtypes of SARS-CoV-2. The proportions of the clusters across the continents revealed distinct geographical distributions. Comprehensive analysis indicated that both genetic factors and human migration factors shaped the specific geographical distribution of the population structure. This study provides a different approach using clustering methods to study the population structure of a never-seen-before and fast-growing species such as SARS-CoV-2. Moreover, clustering techniques can be used for further studies of local population structures of the proliferating virus.

Exploring Unsupervised Machine Learning Classification Methods for Physiological Stress Detection.

Over the past decade, there has been a significant development in wearable health technologies for diagnosis and monitoring, including application to stress monitoring. Most of the wearable stress monitoring systems are built on a supervised learning classification algorithm. These systems rely on the collection of sensor and reference data during the development phase. One of the most challenging tasks in physiological or pathological stress monitoring is the labeling of the physiological signals collected during an experiment. Commonly, different types of self-reporting questionnaires are used to label the perceived stress instances. These questionnaires only capture stress levels at a specific point in time. Moreover, self-reporting is subjective and prone to inaccuracies. This paper explores the potential feasibility of unsupervised learning clustering classifiers such as Affinity Propagation, Balanced Iterative Reducing and Clustering using Hierarchies (BIRCH), K-mean, Mini-Batch K-mean, Mean Shift, Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and Ordering Points To Identify the Clustering Structure (OPTICS) for implementation in stress monitoring wearable devices. Traditional supervised machine learning (linear, ensembles, trees, and neighboring models) classifiers require hand-crafted features and labels while on the other hand, the unsupervised classifier does not require any labels of perceived stress levels and performs classification based on clustering algorithms. The classification results of unsupervised machine learning classifiers are found comparable to supervised machine learning classifiers on two publicly available datasets. The analysis and results of this comparative study demonstrate the potential of unsupervised learning for the development of non-invasive, continuous, and robust detection and monitoring of physiological and pathological stress.

Dynamics analysis of a novel hybrid deep clustering for unsupervised learning by reinforcement of multi-agent to energy saving in intelligent buildings

The heating, ventilating and air conditioning (HVAC) systems energy demand can be reduced by manipulating indoor conditions within the comfort range, which relates to control performance and, simultaneously, achieves peak load shifting toward off-peak hours. Reinforcement learning (RL) is considered a promising technique to solve this problem without an analytical approach, but it has been unable to overcome the awkwardness of an extremely large action space in the real world; it would be quite hard to converge to a set point. The core of the problem with RL is its state space and action space of multi-agent action for building and HVAC systems that have an extremely large amount of training data sets. This makes it difficult to create weights layers accurately of the black-box model. Despite the efforts of past works carried out on deep RL, there are still drawback issues that have not been dealt with as part of the basic elements of large action space and the large-scale nonlinearity due to high thermal inertia. The hybrid deep clustering of multi-agent reinforcement learning (HDCMARL) has the ability to overcome these challenges since the hybrid deep clustering approach has a higher capacity for learning the representation of large space and massive data. The framework of RL agents is a greedy iterative trained and organized as a hybrid layer clustering structure to be able to deal with a non-convex, non-linear and non-separable objective function. The parameters of the hybrid layer are optimized by using the Quasi-Newton (QN) algorithm for fast response signals of agents. That is to say, the main motivation is that the state and action space of multi-agent actions for building HVAC controls are exploding, and the proposed method can overcome this challenge and achieve 32% better performance in energy savings and 21% better performance in thermal comfort than PID.

Student Retention Using Educational Data Mining and Predictive Analytics: A Systematic Literature Review

Student retention is an essential measurement metric in education, indicated by retention rates, which are accumulated as students re-enroll from one academic year to the next. High retention rates can be obtained if institutions aim to provide appropriate support and teaching methods among the various practices to prevent students from deferring their studies. To address this pressing challenge faced by educational institutions, the underlying factors and the methodological aspects of building robust predictive models are reviewed and scrutinized. Educational Data Mining (EDM) and Learning Analytics (LA) have been widely adopted for knowledge discovery from educational data sources, improving the teaching practice, and identifying at-risk students. Various predictive techniques are applied in LA, such as Machine Learning (ML), Statistical Analysis, and Deep Learning (DL). To gain an in-depth review of these techniques, academic publications have been reviewed to highlight their potential to resolve Student Retention issues in education. Additionally, the paper presents a taxonomy of ML approaches and a comprehensive review of the success factors and the features that are not indicative of student performance in three different learning environments: Traditional Learning, Blended Learning, and Online Learning. The survey reveals that supervised ML and DL techniques are broadly applied in Student Retention. However, the application of ensemble and unsupervised learning clustering techniques supporting the heterogenous and homogenous groups of students is generally lacking. Moreover, static and traditional features are commonly used in student performance, ignoring vital factors such as educators-related, cognitive, and personal data. Furthermore, the paper highlights open challenges for future research directions.

Analysis and comparison of three unsupervised learning clustering methods for GNSS multipath signals

Analysis and comparison of three unsupervised learning clustering methods for GNSS multipath signals

Eye-tracking Technology and Data-mining Techniques used for a Behavioral Analysis of Adults engaged in Learning Processes.

Behavioral analysis of adults engaged in learning tasks is a major challenge in the field of adult education. Nowadays, in a world of continuous technological changes and scientific advances, there is a need for life-long learning and education within both formal and non-formal educational environments. In response to this challenge, the use of eye-tracking technology and data-mining techniques, respectively, for supervised (mainly prediction) and unsupervised (specifically cluster analysis) learning, provide methods for the detection of forms of learning among users and/or the classification of their learning styles. In this study, a protocol is proposed for the study of learning styles among adults with and without previous knowledge at different ages (18 to 69-year-old) and at different points throughout the learning process (start and end). Statistical analysis-of-variance techniques mean that differences may be detected between the participants by type of learner and previous knowledge of the task. Likewise, the use ofunsupervised learning clustering techniques throws light on similar forms of learning among the participants across different groups. All these data will facilitate personalized proposals from the teacher for the presentation of each task at different points in the chain of information processing. It will likewise be easier for the teacher to adapt teaching materials to the learning needs of each student or group of students with similar characteristics.

Data-Driven Fault Diagnosis of Chemical Processes Based on Recurrence Plots

A method for the detection and diagnosis of various faults in chemical processes based on the combination of recurrence quantification analysis and unsupervised learning clustering methods is propo...

Mapping study using the unsupervised learning clustering approach

The aim of this research is to analyze the number of criminal cases in Indonesia by utilizing unsupervised learning techniques. The unsupervised learning technique used is data mining by mapping clusters of regions in Indonesia. Sources of data were obtained from the Operations Control Bureau, National Police Headquarters of the Republic of Indonesia through processed data from the Central Statistics Agency (abbreviated as BPS) with data url: https://www.bps.go.id. The data mining method used to map the form of calcter is k-medoid. The data used is data on the number of crimes according to the regional police (2017-2019) which consists of 34 records. The attribute used is the number of crimes in the past three years based on the regional police for each province. The mapping label used is the high cluster (D1) and the low cluster (D2) on the number of criminal acts in Indonesia. The mapping analysis process uses the help of Rapid Miner software. In determining the amount of calcter (k = 2) is done using the Davies Bouldin Index (DBI) parameter with a value of 0.876 (the smaller the better). The results showed that six provinces were in the high cluster (D1) and twenty-eight provinces were in the low cluster (D2). The final centroid in each cluster is 16,008; 21,498; 21,616 (cluster_0 / D1) and 6,994; 7,311; 6,785 (cluster_1 / D2). The six provinces in the high cluster of criminal cases are North Sumatra, South Sumatra, Metro Jaya, West Java, East Java and South Sulawesi. The results of the research are expected to provide information for the government to reduce the number of criminal acts in Indonesia based on the number of clusters that exist.

A Clustering Approach for Jamming Environment Classification

A hierarchical clustering architecture is proposed to deal with the problem of jamming environment classification when multiple noise-like jammers are possibly present. Assuming the availability of clutter-free multichannel data, a two-level hierarchical procedure is devised to unveil the presence of clusters containing range cells experiencing the same jamming interference as the cell under test. Level 1 relies on the use of covariance smoothing and model-order selection rules to make inference on the number of jamming signals affecting each range bin within the radar range swath. Level 2 allows to discriminate among possible different interfering scenarios characterized by the same number of jammers via an unsupervised learning clustering fed by a suitable feature set. At the analysis stage, the performance of the devised architecture is investigated over simulated and measured data (via software-defined radio devices) to highlight the benefits of the approach.

Prediction of human diseases using optimized clustering techniques

In this paper an attempt is made to study the unsupervised learning clustering algorithms such as K-Means, Agglomerative, and Fuzzy C-means Clustering methods. The Nature inspired metaheuristics algorithms such as Salp Swarm Optimization Algorithm (SSOA), Gray Wolf Optimization (GWO), Whale Optimization Algorithm (WOA), Moth-Flame Optimization Algorithm (MFO), BAT Optimization Algorithm (BAT), and Firefly Algorithm (FFA) are used to optimize the unsupervised algorithms for better results. Objective functions like Sum of Squared Errors and Total within-Cluster Variation are taken as objective function in the optimization algorithms. The datasets used in the paper are Heart, Liver and Appendicitis. It has been found that the K-Means have given better results compared to the other algorithms.

ScAIDE: clustering of large-scale single-cell RNA-seq data reveals putative and rare cell types.

Recent advancements in both single-cell RNA-sequencing technology and computational resources facilitate the study of cell types on global populations. Up to millions of cells can now be sequenced in one experiment; thus, accurate and efficient computational methods are needed to provide clustering and post-analysis of assigning putative and rare cell types. Here, we present a novel unsupervised deep learning clustering framework that is robust and highly scalable. To overcome the high level of noise, scAIDE first incorporates an autoencoder-imputation network with a distance-preserved embedding network (AIDE) to learn a good representation of data, and then applies a random projection hashing based k-means algorithm to accommodate the detection of rare cell types. We analyzed a 1.3 million neural cell dataset within 30 min, obtaining 64 clusters which were mapped to 19 putative cell types. In particular, we further identified three different neural stem cell developmental trajectories in these clusters. We also classified two subpopulations of malignant cells in a small glioblastoma dataset using scAIDE. We anticipate that scAIDE would provide a more in-depth understanding of cell development and diseases.

Performance traits of a newly proposed modularity function for spatial networks: Better assessment of clustering for unsupervised learning

The “best” partition of a given network helps in revealing its naturally identifiable structures. The most modular structure is often considered as the best partition. Modularity function, is an objective measure of the quality of partitioning in a given network with that of a random graph (“Null model”), where edge between any two nodes is equally probable, are inappropriate to use for spatially embedded networks. Earlier we have proposed a new modularity function, which does not compare the network with a null model. We have analyzed a 2D and 3D granular networks which can be considered as a spatially embedded network. In all considered systems new method identifies the better partition. New function properly detects the better modular partition in 2D as well as in 3D granular assemblies as compared to the most commonly used modularity function, known as Newman modularity function, and thus is more suitable for unsupervised machine learning.

Empowering the Bandwidth of Continuous-Mode Symmetrical Doherty Amplifiers by Leveraging on Fuzzy Logic Techniques

Achieving a fractional bandwidth (FBW) of more than 60% has been a challenging problem for two-way symmetrical Doherty power amplifiers (TW-SDPAs) that are designed using a continuous-mode technique. As reported in the literature, these designed continuous-mode-based TW-SDPAs possess less than 52% FBW, which cannot satisfactorily meet the challenging, complex, and ever-evolving modulation schemes' demands. To overcome such a limitation, this article proposes a novel approach based on fuzzy logic techniques able to simplify and speed up the design of continuous-mode-based TW-SDPAs with the state-of-the-art FBW. In particular, the proposed technique uses the K -means unsupervised learning clustering algorithm and the continuous-mode technique in a modeled fuzzy logic system environment. As a result, extensive impedance solution design space is readily made available and the optimal impedances required by the carrier and peaking subamplifiers for efficiently operating at the saturation and output-power-back-off (OPBO) levels are automatically obtained. For verification, a TW-SDPA was designed and measured. According to the measured results, the TW-SDPA operates within the 1.2-2.4-GHz frequency band corresponding to 66.7% FBW. As compared with the designed continuous-mode-based TW-SDPAs reported in the literature, this article indicates over 15% increment in FBW. Moreover, 41.59%-81.1% drain efficiency (DE) at saturation, 35%-63% DE at 6-dB OPBO, 42-45-dBm output power, and 7-10.52-dB gain were successfully achieved. Adjacent channel leakage ratio (ACLR) better than -46 dBc and average DE within 46%-55% were successfully recorded after linearization.