Application Of Gaussian Mixture Model Research Articles

Application of Gaussian mixture models to quantify the upper background threshold for perfluorooctane sulfonate (PFOS) in U.S. surface soil.

Studies on the occurrence and environmental distribution of per- and polyfluoroalkyl substances (PFAS) have clearly demonstrated their ubiquity in surface soil as a result of historic and ongoing emissions from various manufacturing and industrial activities worldwide. Given global efforts to characterize and mitigate risk from point source-impacted sites, there is, thus, an urgent need to quantify nonpoint source threshold concentrations (i.e., background) to support site management decisions particularly for perfluorooctane sulfonate (PFOS) as a top priority. Accordingly, this study evaluated the application of Gaussian mixture models (GMMs) fitted to log-transformed PFOS concentrations using nation-wide metadata consisting of thousands of surface soil samples representative of both background and aqueous film-forming foam (AFFF) impacts with unknown proportion. Multiple GMMs were fitted for a given number of components using different methods to account for bias associated with a marginal non-detect fraction (n = 8%) including exclusion, substitution, and imputation. Careful evaluation of the rate of change among multiple goodness-of-fit measures universally justified fitting a 2-component GMM; thus, discriminating between background and AFFF-impacted samples among the metadata. Background threshold PFOS concentrations were defined as the intersection of the probability density functions and ranged between 1.9 and 13.8µg/kg within a broader concentration range extending up to ~ 50,000µg/kg reflecting AFFF impacts. By demonstrating an innovative statistical approach that intelligently incorporates different criteria for model selection, this research makes significant contributions to risk mitigation efforts at point source-impacted sites and lays the groundwork for future targeted regulatory actions.

Gaussian mixture model in clustering acoustic emission signals for characterizing osteoarthritic knees

High-frequency acoustic waves produced by the sudden release of energy within a material caused by mechanical loading, deformation, corrosion, or other forms of damage are detected by Acoustic Emission (AE). It is a non-destructive testing technique utilized to monitor the structural health of materials and components and to detect and locate defects or damage that could lead to failure. Researchers in the biomedical field have lately focused on AE due to its potential for early detection, tracking, and managing various health issues, such as identifying and evaluating joint disorders, tissue damage, etc. However, due to the variance in AE signals produced from complex structures like knee joints, a general statistical method frequently finds it difficult to differentiate between distinct types of AE signals from different kinds of damage. In such instances, machine learning (ML) and deep learning (DL) can be beneficial for differentiating these signals. In this study, the application of Gaussian mixture model (GMM) clustering with principal component analysis (PCA) for dimension reduction was proposed to detect knee osteoarthritis (OA) from AE data. Four groups of participants, Group A (age 20–39), Group B (age 40–59), and Group C (age 60+), along with diagnosed OA as Group D, were included in this study. Avoiding overlapping AE signal parameters from different groups of participants is a significant challenge. GMM was applied to identify and eliminate the coinciding data points to address this issue. Moreover, the impact of this approach on clustering was also investigated in this research, and by utilizing GMM and PCA, compact clustering of the AE data potentially ensured the efficiency and effectiveness of early diagnosis of OA using the AE Technique (AET).

Application of Gaussian Mixture Models for aerosol type analysis in China

The relative contributions of different aerosol types to the total aerosol content were determined using sun photometer derived inversion data from four Aerosol Robotic Network (AERONET) stations in China, i.e. Beijing, NUIST, Hong Kong PolyU and SACOL. Five main aerosol types were considered: dust (DU), mixed (MI), urban industry (UI), biomass burning (BB) and maritime (MA). The identification of these aerosol types is based on cluster analysis using Gaussian mixture models (GMMs), together with the Mahalanobis distance (MD) to each reference centroid. The results show that the aerosol types are distinctly different between the different sites as regards fine and coarse particles. MI is the dominant aerosol type at four locations, with the largest contribution at NUIST (73.4%). The largest contribution of DU (18.8%) is observed at SACOL and at Beijing the contribution of UI is largest (33.7%). BB contributions are relatively small, with the largest at NUIST (2.7%). For the analysis, two periods are considered, i.e. before and after the implementation of the national ambient air quality standard (GB3095–2012) in 2012. The clustering results revealed a changing trend in the proportion of UI and MI aerosols in Chinese cities. Although the proportion of UI aerosols has increased, MI aerosols still remain dominant. The proportion of DU in SACOL station also increased. The seasonal analysis shows that DU dominates in spring at the SACOL station (55.3%), which is near the dust source area. Among the other stations, the contribution of DU due to long-range transport from the source regions, is largest (10.6%) in Beijing, with a small contribution at NUIST. The aerosol volume size distributions in different regions exhibit substantial differences due to variations in aerosol types: Beijing and SACOL are dominated by coarse mode particles, while NUIST and Hong Kong PolyU are dominated by fine mode particles.

Dynamically recrystallized grains identified via the application of Gaussian mixture model to EBSD data

Crystals formed during ductile deformation known as the recrystallized grains have been revealed to provide syn-deformational stress and strain information. Despite its importance, current methods of distinguishing recrystallized grains are subjective and lack consideration to the evolving microstructures. Using data obtained from electron backscattered diffraction mapping, we provide an unsupervised method to estimate the recrystallized fraction and the recrystallized grain size. The Gaussian mixture model was utilized to upon the grain orientation spread that measures intragranular lattice distortion to separate the grains into clusters. The counterpart of the recrystallized grain, the relict grains which are grains present prior to deformation are identified as grains distributed under the cluster with the largest mean grain orientation spread. Three zones could be separated from the Gaussian mixture model results representing the recrystallized zone, mixture zone and relict zone. Grains located within the mixture zone have an innate probability to be classified as either relict or recrystallized grains. The application of the Monte Carlo approximation enables the evaluation of error for the recrystallized fraction and grain size. Comparison to previously available methods shows that the current method avoids ambiguity in the selection of the cut-off threshold when identifying the recrystallized grains.

Grocery Apps and Consumer Purchase Behavior: Application of Gaussian Mixture Model and Multi-Layer Perceptron Algorithm

The purpose of this study is to investigate and compare the popularity of common grocery apps in Hungary as well as Iran. The data were gathered from Iranian and Hungarian users who had at least one online purchase experience using a grocery app. A Gaussian mixture model (GMM) and multi-layer perceptron (MLP) are used as supervised and unsupervised machine learning algorithms with Python programming to cluster customers and predict consumer behavior. The results revealed that Wolt in Hungary and Snappfood in Iran are the most popular grocery apps. Users in Iran are divided into three groups of users of app services and the type of full covariance has higher accuracy compared to the other three types (96%). Meanwhile, we found that the five apps used in Hungary have provided 95% accuracy from the users’ point of view based on the diagonal covariance. The MSE value (overfitting and cross-validation) is less than 0.1 in the MLP algorithm, which shows an acceptable amount of error. The results of overfitting indicate the proper fit of the MLP model. The findings of this study could be important for managers of online businesses. In the clustering section, the accuracy and value of consumer demographic information have been emphasized. Additionally, in the classification and prediction section, a kind of “customization” has been performed with an emphasis on market segmentation. This research used GMM and MLP machine learning algorithms as a creative way to cluster and classify consumers.

Application of Gaussian Mixture Model for the Analysis of the Nanoindentation Test Results of the Metakaolin-based Geopolymer with Different Silicon-to-Aluminum Molar Ratio

Application of Gaussian Mixture Model for the Analysis of the Nanoindentation Test Results of the Metakaolin-based Geopolymer with Different Silicon-to-Aluminum Molar Ratio

Application of Gaussian Mixture Model and Geostatistical Co-simulation for Resource Modeling of Geometallurgical Variables

This work addresses the practice of resource calculation for geometallurgical variables. Similar to mineral resource modeling, estimation domains for geometallurgical variables should be identified first. Then, the geometallurgical variables that are deemed homogeneous need to be modeled separately in each domain. A difficulty for this is related to the geometallurgical variables that can rarely be in agreement with the geological interpretation of a deposit. To circumvent this difficulty, a machine learning approach, namely Gaussian mixture model technique, is employed in an Fe ore deposit to obtain clusters that can turn out the geometallurgical domains. After corroborating that the obtained domains are reasonable from a geometallurgical perspective, a technique is provided to co-simulate the geometallurgical variables within the attained geometallurgical domains following a probabilistic cascade approach. The latter allows incorporation of cross-dependency among the variables that usually are neglected in the modeling process. The algorithm showed that the proposed technique is statistically valid and can be applied for optimum ore processing plant and strategic mine design, where defining the grade alone may not be enough for deciding on further optimization of a mining project. It is also showed that as an instruction, how the proposed algorithm can provide the recovery functions of the geometallurgical variables for resource calculation.

Application of Gaussian Mixture Model (GMM) Cluster Analysis on Morphological Data of Archaeological Human Remains

Application of Gaussian Mixture Model (GMM) Cluster Analysis on Morphological Data of Archaeological Human Remains

The application of Gaussian mixture models for signal quantification in MALDI-TOF mass spectrometry of peptides.

Matrix assisted laser desorption/ionization time-of-flight (MALDI-TOF) coupled with stable isotope standards (SIS) has been used to quantify native peptides. This peptide quantification by MALDI-TOF approach has difficulties quantifying samples containing peptides with ion currents in overlapping spectra. In these overlapping spectra the currents sum together, which modify the peak heights and make normal SIS estimation problematic. An approach using Gaussian mixtures based on known physical constants to model the isotopic cluster of a known compound is proposed here. The characteristics of this approach are examined for single and overlapping compounds. The approach is compared to two commonly used SIS quantification methods for single compound, namely Peak Intensity method and Riemann sum area under the curve (AUC) method. For studying the characteristics of the Gaussian mixture method, Angiotensin II, Angiotensin-2-10, and Angiotenisn-1-9 and their associated SIS peptides were used. The findings suggest, Gaussian mixture method has similar characteristics as the two methods compared for estimating the quantity of isolated isotopic clusters for single compounds. All three methods were tested using MALDI-TOF mass spectra collected for peptides of the renin-angiotensin system. The Gaussian mixture method accurately estimated the native to labeled ratio of several isolated angiotensin peptides (5.2% error in ratio estimation) with similar estimation errors to those calculated using peak intensity and Riemann sum AUC methods (5.9% and 7.7%, respectively). For overlapping angiotensin peptides, (where the other two methods are not applicable) the estimation error of the Gaussian mixture was 6.8%, which is within the acceptable range. In summary, for single compounds the Gaussian mixture method is equivalent or marginally superior compared to the existing methods of peptide quantification and is capable of quantifying overlapping (convolved) peptides within the acceptable margin of error.

Sequential Monte Carlo implementation and state extraction of the group probability hypothesis density filter for partly unresolvable group targets-tracking problem

This study proposes a sequential Monte Carlo (SMC) implementation to the Mahler's group probability hypothesis density filter (PHDF) for partly unresolvable group targets tracking problem. A potential limitation of the group PHDF is that it cannot be used to determine the number of groups. Therefore we have to jointly extract the group number and states from the proposed group SMC-PHDF at each time step. We propose to fit the resampled particles of the group SMC-PHDF by application of Gaussian mixture models with unknown component number. In the mixture, the number and parameters of the components correspond to the number and states of the groups over the observation region. The Markov chain Monte Carlo (MCMC) algorithm is proposed to estimate the component parameters of the mixture. The estimate of component number of the mixture can be derived by a component management strategy. In simulation, the proposed group SMC-PHDF with the expectation maximum (EM) and MCMC extractions are, respectively, used to detect and track the groups. Hundred Monte Carlo simulation results show that the latter outperforms the former a lot in estimating the group number and states, although the computational requirement of the MCMC extraction is more expensive than the EM extraction.

Comparative Study of Gaussian Mixture Model and Radial Basis Function for Voice Recognition

A comparative study of the application of Gaussian Mixture Model (GMM) and Radial Basis Function (RBF) in biometric recognition of voice has been carried out and presented. The application of machine learning techniques to biometric authentication and recognition problems has gained a widespread acceptance. In this research, a GMM model was trained, using Expectation Maximization (EM) algorithm, on a dataset containing 10 classes of vowels and the model was used to predict the appropriate classes using a validation dataset. For experimental validity, the model was compared to the performance of two different versions of RBF model using the same learning and validation datasets. The results showed very close recognition accuracy between the GMM and the standard RBF model, but with GMM performing better than the standard RBF by less than 1% and the two models outperformed similar models reported in literature. The DTREG version of RBF outperformed the other two models by producing 94.8% recognition accuracy. In terms of recognition time, the standard RBF was found to be the fastest among the three models.

On-line statistical monitoring of batch processes using Gaussian mixture model

The statistical monitoring of batch manufacturing processes is considered. It is known that conventional monitoring approaches, e.g. principal component analysis (PCA), are not applicable when the normal operating conditions of the process cannot be sufficiently represented by a Gaussian distribution. To address this issue, Gaussian mixture model (GMM) has been proposed to estimate the probability density function of the process nominal data, with improved monitoring results having been reported for continuous processes. This paper extends the application of GMM to on-line monitoring of batch processes, and the proposed method is demonstrated through its application to a batch semiconductor etch process.

Probability distributions and leveraged trading strategies: an application of Gaussian mixture models to the Morgan Stanley Technology Index Tracking Fund

The purpose of this paper is twofold. Firstly, to assess the merit of estimating probability density functions rather than level or direction forecasts for one-day-ahead forecasts of the Morgan Stanley Technology Index Tracking Fund (MTK). This is implemented using a Gaussian mixture model neural network, benchmarking the results against standard forecasting models, namely a naïve model, a moving average convergence divergence technical model (MACD), an autoregressive moving average model (ARMA), a logistic regression model (LOGIT) and a multi-layer perceptron network (MLP). Secondly, we examine the possibilities of improving the trading performance of those models with confirmation filters and leverage. While the two network models outperform all of the benchmark models, the Gaussian mixture model does best: it is worth noting that it does well on a time series where the training period is showing a strong uptrend while the out-of-sample period is characterized by a downtrend.

Probability distributions, trading strategies and leverage: an application of Gaussian mixture models

AbstractThe purpose of this paper is twofold. Firstly, to assess the merit of estimating probability density functions rather than level or classification estimations on a one‐day‐ahead forecasting task of the EUR/USD time series. This is implemented using a Gaussian mixture model neural network, benchmarking the results against standard forecasting models, namely a naïve model, a moving average convergence divergence technical model (MACD), an autoregressive moving average model (ARMA), a logistic regression model (LOGIT) and a multi‐layer perceptron network (MLP). Secondly, to examine the possibilities of improving the trading performance of those models with confirmation filters and leverage.While the benchmark models perform best without confirmation filters and leverage, the Gaussian mixture model outperforms all of the benchmarks when taking advantage of the possibilities offered by a combination of more sophisticated trading strategies and leverage. This might be due to the ability of the Gaussian mixture model to identify successfully trades with a high Sharpe ratio. Copyright © 2004 John Wiley & Sons, Ltd.