Classical K-means Research Articles

Emulation of high-performance correlation-based quantum clustering algorithm for two-dimensional data on FPGA

Clustering algorithms are used to classify the unlabeled data into a number of categories with polynomial time complexity. Quantum clustering algorithms are developed to improve the performance and to achieve higher gain. In this work, we implement the quantum k-means clustering algorithm on field-programmable gate array (FPGA) by exploiting the implicit parallelism of the FPGA technology to achieve high speed among the software-implemented recent proposals. To do that, we establish a new method to measure the inner product between two qubits which is based on the correlation between the Euclidean distance and the inner product. We also optimize the quantum gates in terms of speed and removing the discretization error. Experimental results show a reduction in the running time by 500× as compared to the classical k-means algorithm for the A1 standard dataset.

Students Activity Recognition by Heart Rate Monitoring in Classroom using K-Means Classification

Background: Heartbeat playing the main roles in our life. With the heartbeat, the anxiety level can be known. Most of the heartbeat is used in the exercise. Heart rate measurement is unique and uncontrollable by any human being.Objective: This research aims to learn student’s actions by monitoring the heart rate. In this paper, we are measuring the student reaction and action in classroom can give impact on teacher’s way of delivery when in the teaching session. In monitoring, student’s behavior may give feedback whether the teaching session have positive or negative outcome.Methods: The method we use is K-Means algorithm. Firstly, we need to know the student’s normal heartbeat as benchmark. We used Hexiware for collecting data from students’ hear beat. We perform the classification where K is benchmark students’ heartbeat. K-Means algorithm performs classification of the heart rate measurement of students.Results: We did the testing for five students in different subjects. It shows that all students have anxiety during the testing and presentation. Its consistency because we tested 5 students with mixes activities in the classroom, where the student has quiz, presentation and only teaching.Conclusion: Heart rate during studying in the classroom can change the education world in improving the efficiency of knowledge transfer between student and teacher. This research may act as basic way in monitoring student behavior in the classroom. We have tested for 5 students. Three students have their anxiety in classroom during the exam, presentation, and question. Two students have normal rate during the seminar and lecturer. The drawback, Hexiware is capturing average of ten minutes and tested in different classes and students. In future, we need just measure one student for all the subjects and Hexiware need to configure in one minute.

Collaborative Graph Convolutional Networks: Unsupervised Learning Meets Semi-Supervised Learning

Graph convolutional networks (GCN) have achieved promising performance in attributed graph clustering and semi-supervised node classification because it is capable of modeling complex graphical structure, and jointly learning both features and relations of nodes. Inspired by the success of unsupervised learning in the training of deep models, we wonder whether graph-based unsupervised learning can collaboratively boost the performance of semi-supervised learning. In this paper, we propose a multi-task graph learning model, called collaborative graph convolutional networks (CGCN). CGCN is composed of an attributed graph clustering network and a semi-supervised node classification network. As Gaussian mixture models can effectively discover the inherent complex data distributions, a new end to end attributed graph clustering network is designed by combining variational graph auto-encoder with Gaussian mixture models (GMM-VGAE) rather than the classic k-means. If the pseudo-label of an unlabeled sample assigned by GMM-VGAE is consistent with the prediction of the semi-supervised GCN, it is selected to further boost the performance of semi-supervised learning with the help of the pseudo-labels. Extensive experiments on benchmark graph datasets validate the superiority of our proposed GMM-VGAE compared with the state-of-the-art attributed graph clustering networks. The performance of node classification is greatly improved by our proposed CGCN, which verifies graph-based unsupervised learning can be well exploited to enhance the performance of semi-supervised learning.

Feature based clustering technique for investigation of domestic load profiles and probabilistic variation assessment: Smart meter dataset

Abstract Dimensions of electrical distribution network datasets have been increasing exponentially as result of global acceptance of toward smart metering projects for secure implementation of demand response strategies and attaining satisfactory operation of electrical distribution network. Traditional approaches of analyzing these datasets have often been prone to losing of important information for instance averaging and aggregating of data; such loss of information can prove imperative in this era of demand side management and demand response. High dimensionality of distribution dataset is prominent factor for popularity of such conventional perspective toward large datasets. However, recent evolution in data mining have tossed various dimensionality reduction techniques expressing minimal loss of information. This paper proposes a feature based clustering algorithm aimed at dimensionality reduction, load profile characterization and probabilistic load variation assessment as a case study for smart village project of Nana Kajaliyala village, Gujarat, India. Proposed algorithm attains profile characterization using classical k-means alongside an empirical feature selection countering high dimensionality. A comparative evaluation of proposed algorithm with other popular techniques like self-organizing map (SOM) and classical k-means is presented in this paper. Moreover, a novel probabilistic analysis approach is conferred, which is directed at assessment of load variation, peak risk analysis of individual consumers. Determined statistical assessment measures in this paper would aid the utility with capability to execute cognitive decision making and reduce aggregate technical and commercial losses. Furthermore, load labels assigned to each characteristic profile could help managing load requirements, and planning future operations.

Penerapan Metode K-Means Dan C4.5 Untuk Prediksi Penderita Diabetes

The purpose of this study is to predict HbA1c in diabetics. The obstacles behind the prediction of HbA1c is the limitations in the laboratory to provide services for diabetics regarding HbA1 check-up. HbA1c prediction is made by a combination of K-Means and C4.5 methods. K-Means is used to classify continuous data. From the results of the K-Means classification will be used by C4.5 to create a rule (decision tree). The prediction results obtained will be carried out as a validation process to determine the level of accuracy by using K-Fold Cross-Validation. The accuracy value obtained is 72%. The resulting benefit from the prediction of HbA1c can be used as an alternative solution to overcome limitations in the laboratory in terms of HbA1c check-up servicing and the results of HbA1c prediction can also be used as a recommendation by doctor in determining the medical decision for diabetics.

An expert system as an applied tool for the conservation of semi-natural grasslands on the Iberian Peninsula

Semi-natural grasslands have an important conservation value in Europe but have been in continuous decline since the early twentieth century. Valuable and well-preserved semi-natural grasslands must be identified in order to plan conservation and restoration strategies. The main aim of this paper is to develop an expert system based on diagnostic species for the identification and assessment of EU semi-natural grasslands using a dataset of 3238 vegetation plots from the Iberian Peninsula (Portugal and Spain, including the French Pyrenees). The dataset was obtained by stratified resampling of a larger dataset of 12,204 vegetation plots containing other related grasslands and herbaceous vegetation types. We applied a k-means classification on the resampled dataset and selected the diagnostic species for each group of EU semi-natural grasslands in order to develop an expert classification system with definitions based on their species composition and biogeographic distribution. The expert-based classification was applied to 14 vegetation subtypes belonging to semi-natural grasslands in the Habitats Directive, corresponding to semi-dry calcareous grasslands (Habitat 6210), Nardus grasslands (6230), hay meadows (6510, 6520) and Molinia meadows on peaty soils (6410). We provide a practical and easy tool for the identification, monitoring and assessment of EU semi-natural grasslands on the Iberian Peninsula.

An approximation algorithm for the uniform capacitated k-means problem

In this paper, we consider the uniform capacitated k-means problem (UC-k-means), an extension of the classical k-means problem (k-means) in machine learning. In the UC-k-means, we are given a set $$\mathcal {D}$$ of n points in d-dimensional space and an integer k. Every point in the d-dimensional space has an uniform capacity which is an upper bound on the number of points in $$\mathcal {D}$$ that can be connected to this point. Every two-point pair in the space has an associated connecting cost, which is equal to the square of the distance between these two points. We want to find at most k points in the space as centers and connect every point in $$\mathcal {D}$$ to some center without violating the capacity constraint, such that the total connecting costs is minimized. Based on the technique of local search, we present a bi-criteria approximation algorithm, which has a constant approximation guarantee and violates the cardinality constraint within a constant factor, for the UC-k-means.

Research on Parallel Adaptive Canopy-K-Means Clustering Algorithm for Big Data Mining Based on Cloud Platform

Firstly, this paper introduces the types of clustering algorithm, and introduces the classical K-means algorithm and canopy algorithm in detail. Then, combining the map reduce computing model and spark cloud computing framework, this paper introduces the parallel Canopy-K-means algorithm after using Canopy algorithm to optimize the initial value of K-means algorithm. However, because Canopy algorithm needs to introduce a new distance threshold parameter T2, and the parameter needs to be set by human experience, it is difficult to determine the parameter artificially for large data, so this paper proposes a parallel adaptive Canopy-K-means algorithm, which can be used in cloud computing framework to determine the distance threshold parameter T2 adaptively based on statistical method. Using the parallelism of Map-Reduce computing model, the parallel Canopy-K-means algorithm is optimized by adaptive parameter estimation, which solves the problem that parameters depend on manual experience selection in Canopy process. After introducing the relevant theories and derivation process of this algorithm, cloud computing experiment platform is built based on the Spark framework, and the contrast experiments were performed using the Stanford Large Network Dataset Collection (SNAP) dataset and self-built Dimension Networks dataset. The experimental results show that the proposed method is effective.

Modelling hydrological responses under climate change using machine learning algorithms – semi-arid river basin of peninsular India

AbstractCatchment scale conceptual hydrological models apply calibration parameters entirely based on observed historical data in the climate change impact assessment. The study used the most advanced machine learning algorithms based on Ensemble Regression and Random Forest models to develop dynamically calibrated factors which can form as a basis for the analysis of hydrological responses under climate change. The Random Forest algorithm was identified as a robust method to model the calibration factors with limited data for training and testing with precipitation, evapotranspiration and uncalibrated runoff based on various performance measures. The developed model was further used to study the runoff response under climate change variability of precipitation and temperatures. A statistical downscaling model based on K-means clustering, Classification and Regression Trees and Support Vector Regression was used to develop the precipitation and temperature projections based on MIROC GCM outputs with the RCP 4.5 scenario. The proposed modelling framework has been demonstrated on a semi-arid river basin of peninsular India, Krishna River Basin (KRB). The basin outlet runoff was predicted to decrease (13.26%) for future scenarios under climate change due to an increase in temperature (0.6 °C), compared to a precipitation increase (13.12%), resulting in an overall reduction in water availability over KRB.

Habitat choices of African buffalo (Syncerus caffer) and plains zebra (Equus quagga) in a heterogeneous protected area

Abstract ContextAn understanding of large herbivore habitat choices in heterogeneous African protected areas is important for the better management of these key ecosystems. AimsTo determine habitat use of African buffalo (Syncerus caffer) and plains zebra (Equus quagga) in a heterogeneous protected area. MethodsZambezi National Park (ZNP), Zimbabwe, was divided into five vegetation types using an unsupervised classification on a Landsat satellite image that was classified into five land cover classes, using the K-means classification algorithm. African buffalo and plains zebra densities were then determined in each vegetation type using road transect surveys monthly between January 2013 and December 2015. Normalised difference vegetation index (NDVI), grass biomass, grass height and grass quality (nitrogen, calcium, phosphorus and acid detergent fibre content) were determined in each vegetation type during the wet (November to April) and dry (August to October) seasons to establish their quality as habitats for African buffalo and plains zebra. Key resultsBoth African buffalo and plains zebra mostly foraged in mixed and grassland areas, and avoided Zambezi teak vegetation type. Zambezi teak vegetation type had high NDVI due to the dense tree cover. Both African buffalo and plains zebra preferred vegetation types with intermediate grass biomass (approximately 300gm−2) and grass height (approximately 16cm). Grass nutritive value (in terms of nitrogen, phosphorus and acid detergent fibre) declined from wet to dry season in all vegetation types. ConclusionsAfrican buffalo and plains zebra in the ZNP confined their habitat use mostly to two vegetation types (mixed and grassland), which together covered 25% of the protected area. ImplicationsTeak (Baikiaea plurijuga) vegetation, which accounted for about 60% of the ZNP, was avoided by both African buffalo and plains zebra, suggesting that a significant part of the protected area was not used by the two herbivores.

Cluster Density Properties Define a Graph for Effective Pattern Feature Selection

Feature selection is a challenging problem that occurs in the high-dimensional data analysis of many major applications. It addresses the curse of dimensionality by determining a small set of features to represent high-dimensional data without significant or noticeable loss of information. The purpose of this study is to develop and investigate a new unsupervised feature selection method which uses the k-influence space concept and subspace learning to map features onto a weighted graph and rank them by importance according to the PageRank graph centrality measure. The graph design in this method promotes feature relevance, downgrades redundancy, and is robust to outliers and cluster imbalances. In K-Means classification experiments using the ASU feature selection testing datasets, the method produces better accuracy and normalized mutual information results than state-of-the-art unsupervised feature selection algorithms. In a further evaluation, using a dataset of over 14,000 tweets, conventional classification of features selected by the method gave better sentiment analysis results than deep learning feature selection and classification.

Prediction of the Corn Grains Yield through Artificial Intelligence

Abstract Currently, the determination of the quality of the cereals is done manually by grain classifier experts prior to the marketing stage. In this paper we present a web software tool that allows determining the quality level of a corn sample automatically from an image of it. Image processing algorithms were implemented to correct distortions caused mainly by the capture process. The K-Means classification algorithm was used and a function was developed to calculate the hectolitre weight in relation to the sample area. The results obtained by the application for grades 1 and 2, are close to those measured by the experts. However, those for grade 3 have not been similar since the subsamples selected were not representative.

A Study of Inductive Biases for Unsupervised Speech Representation Learning

Distributed representations, or embeddings, are commonly learned without supervision on very large unannotated corpora for natural language processing. In speech processing, deep network-based representations such as bottlenecks and x-vectors have had some success,but are limited to supervised or partly supervised settings where annotations are available and are not optimized to separate underlying factors. Here, we propose a generative model with deep encoders and decoders that can learn interpretable speech representations without supervision. Our inductive biases operate as prior distributions in a variational autoencoder model and allow us to separate several latent variables along a continuous range of time-scale properties, as opposed to binary oppositions or hierarchical factorization that have been previously proposed. On simulated data, we confirm that these biases enable the model to accurately recover phonetic and speaker underlying factors. On TIMIT and LibriSpeech, they yield representations that separate phonetic and speaker information, as evidenced by unsupervised results on downstream phoneme and speaker classification tasks using a simple k-means classifier.

Assessment of pre-treatment and classification methods for Java paddy field cropping pattern detection on MODIS images

Intensive paddy cropping in Java island has an intricate pattern both in spatial and temporal distribution due to landscape and scheduling complexity, in this case the temporal pattern is a key factor in predicting the paddy cropping season. Cropping pattern relevant to the temporal vegetation indices which can be obtained from MODIS temporal imagery. Monitoring of the cropping pattern is a critical factor in preserving the national food security of Indonesia. Remote sensing can be used as an alternative solution to monitor paddy cropping pattern by incorporating statistics analysis and signal processing methods into pre-processing and classification techniques. The main focus of this research is to evaluate the combination of those techniques implemented in MODIS images to classify the annual cropping patterns of paddy crop in Java island. The evaluated techniques involve combinations of four parameters including (1) temporal reconstruction (R0: untreated; R1: linear interpolation), (2) noise reduction (F0: untreated; F1: band-pass FFT; F2: wavelet filter), (3) unsupervised k-means classification using distance parameters (D1: Euclidean; D2: Mahalanobis), and (4) k number of classes (64, 128, 256). The classification validation was evaluated based on the paddy field site visit in several locations across Java island from the years of 2014 until 2017. The best combination (R1-F1-D1-256, the score of 74.5) was determined by the average score calculated from overall classification accuracy, percentage normality of class member, and the average correlation of class member. This combination can describe the highest number of cropping patterns (i.e., 26 patterns) that cover four types of paddy cropping systems (i.e., non-paddy, single paddy cropping, double paddy cropping, and triple paddy cropping). This selected technique can also be used to calculate the annual paddy cropping area.

Two-Stage Optimization Model for Two-Side Daily Reserve Capacity of a Power System Considering Demand Response and Wind Power Consumption

Today, wind power is becoming an important energy source for the future development of electric energy due to its clean and environmentally friendly characteristics. However, due to the uncertainty of incoming wind, the utilization efficiency of wind energy is extremely low, which means the problem of wind curtailment becomes more and more serious. To solve the issue of wind power large-scale consumption, a two-stage stochastic optimization model is established in this paper. Different from other research frameworks, a novel two-side reserve capacity mechanism, which simultaneously takes into account supply side and demand side, is designed to ensure the stable consumption of wind power in the real-time market stage. Specifically, the reserve capacity of thermal power units is considered on the supply side, and the demand response is introduced as the reserve capacity on the demand side. At the same time, the compensation mechanism of reserve capacity is introduced to encourage generation companies (GENCOs) to actively participate in the power balance process of the real-time market. In terms of solution method, compared with the traditional k-means clustering method, this paper uses the K-means classification based on numerical weather prediction (K-means-NWP) scenario clustering method to better describe the fluctuation of wind power output. Finally, an example simulation is conducted to analyze the influence of reserve capacity compensation mechanism and system parameters on wind power consumption results. The results demonstrate that with the introduction of reserve capacity compensation mechanism, the wind curtailment quantity of the power system has a significant reduction. Besides, the income of GENCOs is gradually increasing, which motivates their enthusiasm to provide reserve capacity. Furthermore, the reserve capacity mechanism designed in this paper promotes the consumption of wind power and the sustainable development of renewable energy.

The “Gene Cube”: A Novel Approach to Three-dimensional Clustering of Gene Expression Data

Background:A very popular technique for isolating significant genes from cancerous tissues is the application of various clustering algorithms on data obtained by DNA microarray experiments.Aim:The objective of the present work is to take into consideration the chromosomal identity of every gene before the clustering, by creating a three-dimensional structure of the form Chromosomes×Genes×Samples. Further on, the k-Means algorithm and a triclustering technique called δ- TRIMAX, are applied independently on the structure.Materials and Methods:The present algorithm was developed using the Python programming language (v. 3.5.1). For this work, we used two distinct public datasets containing healthy control samples and tissue samples from bladder cancer patients. Background correction was performed by subtracting the median global background from the median local Background from the signal intensity. The quantile normalization method has been applied for sample normalization. Three known algorithms have been applied for testing the “gene cube”, a classical k-means, a transformed 3D k-means and the δ-TRIMAX.Results:Our proposed data structure consists of a 3D matrix of the form Chromosomes×Genes×Samples. Clustering analysis of that structure manifested very good results as we were able to identify gene expression patterns among samples, genes and chromosomes. Discussion: to the best of our knowledge, this is the first time that such a structure is reported and it consists of a useful tool towards gene classification from high-throughput gene expression experiments.Conclusion:Such approaches could prove useful towards the understanding of disease mechanics and tumors in particular.

Enhanced Classification of Service Usages with Human Trajectory Data for Location Recommendation Systems

The rapid growth of mobile messaging apps has led to an important process to manage social networks based on the localization of internet traffic in different types of use of in-app services. In the past researches, Improved Multi-Context Trajectory Embedding Model with Service Usage Classification Method (IMC-TEM-SUCM) has been proposed to recommend the locations based on the trajectory data of individuals and their service usage types. In this model, the traffic features were classified by using Random Forest (RF) classifier whereas the outlier was detected by clustering Hidden Markov Model (HMM). However, the RF was supervised classifier which requires knowledge about the class label of data. Also, a huge amount of data was needed to train a clustering HMM. Therefore, in this article, an IMC-TEM with Enhanced SUCM (IMC-TEM-ESUCM) is proposed in which an unsupervised classifier, namely K-means clustering is proposed to classify the service usage types. Initially, traffic flows are split into different sessions and dialogs using a combined hierarchical clustering and thresholding heuristics technique. Then, the traffic features are extracted based on the packet length and time delay. After that, K-means classification is proposed to classify the service usage types and also DBSCAN is proposed to detect the outliers. Finally, the experimental results on two different datasets show that the proposed model achieves higher performance than the existing model in terms of precision, recall, f-measure and accuracy.

A Signature Verification System with Ensemble Classifier

Handwritten signature is considered as one of the established authentication process to study the behavioral nature of a person. This paper focuses on verification of offline handwritten signatures (for English scripts) as either genuine or forgery. Here the considered samples are genuine, skilled and simple forgeries. The verification is carried out by ensembling the three base classifiers Naive Bayes (NB), K-Nearest Neighbor (KNN) and Kmeans classifiers. The accuracies for skilled and simple forgeries are obtained as 86 % and 92 % respectively.

Wind Generation Impact in Transmission Expansion Planning

This research proposes a post-stage investment analysis in the transmission expansion network planning problem considering the intermittence of renewable energy sources, especially wind power sources. The main goal of this study is to analyze the voltage profiles in all the system buses and the level of electrical losses in the lines using AC flow for impact evaluation of renewable energy insertion. This research also performs an additional analysis considering wind energy insertion during the planning stage. This analysis is to demonstrate the effect in the stage of investment in transmission lines. The variability of these sources is represented through scenarios obtained by the K-means classification algorithm that allows the preservation of the correlation among generating stations. The methodology performance is tested using the IEEE 24-bus system, which is modified to include a significant share of wind energy.

Using Clustered Heat Maps in Mineral Exploration to Visualize Volcanic-Hosted Massive Sulfide Alteration and Mineralization

This study proposes an extension of a visualization approach common in biochemistry (the clustered heat maps—CHMs) to geochemical data with the main objective of detecting hydrothermal alteration and mineralization. The approach allows superior visualization of unsupervised cluster analysis results. We consider two examples: a synthetic case study and an application to public data derived from the Canadian Flin Flon volcanic-hosted massive sulfide deposits. A series of experiments were run on a synthetic dataset with the aim of understanding the effect of noise and how random data sampling of variable specimen population size influences results of a variety of clustering algorithms (including K-means and other hierarchical methods) and their visualization using CHMs. These experiments on synthetic data provided the basis to propose a possible workflow for the selection of optimal classifiers to be applied on natural data and the definition of an appropriate parametrization (distance metrics and clustering algorithm). Natural data analysis provides direct evidence of how CHMs can be a fruitful approach in mineral exploration if compared to other cluster analysis methods (e.g., classic K-means or hierarchical methods), CHMs provide the opportunity of examining an additional dimension of clustering and still view chemical compositions (although in a transformed space) in a single plot. Facilitated selection of appropriate levels of granularity (G), which regulates the scale of clustering in a CHM, was found to be an instrumental tool and led to the successful separation of clusters representative of major lithological transitions vs. smaller clusters, at higher granularity, isolating VHMS alteration and mineralization. Integration of statistical tests conducted on synthetic data, together with CHM’s visualization of the classification results led us to consider the Manhattan–Ward classifier as an optimal pair for the Flin Flon dataset, despite its limitations induced by the ‘uniform effect.’