Machine Learning Clustering Techniques Research Articles

Cryptocurrency Turmoil: Unraveling the Collapse of a Unified Stablecoin (USTC) through Twitter as a Passive Sensor.

This study sought to explore whether Twitter, as a passive sensor, could have foreseen the collapse of the Unified Stablecoin (USTC). In May 2022, in just a few days, the cryptocurrency went to near-zero valuation. Analyzing 244,312 tweets from 89,449 distinct accounts between April and June 2022, this study delved into the correlation between personal sentiments in tweets and the USTC market value, revealing a moderate correlation with polarity. While sentiment analysis has often been used to predict market prices, the results suggest the challenge of foreseeing sudden catastrophic events like the USTC collapse solely through sentiment analysis. The analysis uncovered unexpected global interest and noted positive sentiments during the collapse. Additionally, it identified events such as the launch of the new Terra blockchain (referred to as "Terra 2.0") that triggered positive surges. Leveraging machine learning clustering techniques, this study also identified distinct user behaviors, providing valuable insights into influential figures in the cryptocurrency space. This comprehensive analysis marks an initial step toward understanding sudden and catastrophic phenomena in the cryptocurrency market.

Energy efficient power cap configurations through Pareto front analysis and machine learning categorization

AbstractThe growing demand for more computing resources has increased the overall energy consumption of computer systems. To support this increasing demand, power and energy consumption must be considered as a constraint on software execution. Modern architectures provide tools for managing the power constraints of a system directly. The Intel Power Cap is a relatively new tool developed to give users fine-grained control over power usage at the central processing unit (CPU) level. The complexity of these tools, in addition to the high variety of modern heterogeneous architectures, hinders predictions of the energy consumption and the performance of any target software. The application of power capping technologies usually leads to the bi-objective optimization problem for energy efficiency and execution time but optimal power constraints could also produce exceeding performance losses. Thus, methods and tools are needed to calculate the proper parameters for power capping technologies, and to optimize energy efficiency. We propose a methodology to analyze the performance and the energy efficiency trade-offs using this power cap technology for a given application. A Pareto front is extracted for the multi-objective performance and energy problem, which represents multiple feasible configurations for both objectives. An extensive experimentation is carried out to categorize the different applications to determine the overall optimal power cap configurations. We propose the use of machine learning (ML) clustering techniques to categorize each application in the target architecture. The use of ML allows us to automate the process and simplifies the effort required to solve the optimization problem. A practical case is presented where we categorize the applications using ML techniques, with the possibility of adding a new application into an existing categorization.

Statistical properties of cerebral near infrared and intracranial pressure-based cerebrovascular reactivity metrics in moderate and severe neural injury: a machine learning and time-series analysis

BackgroundCerebrovascular reactivity has been identified as a key contributor to secondary injury following traumatic brain injury (TBI). Prevalent intracranial pressure (ICP) based indices of cerebrovascular reactivity are limited by their invasive nature and poor spatial resolution. Fortunately, interest has been building around near infrared spectroscopy (NIRS) based measures of cerebrovascular reactivity that utilize regional cerebral oxygen saturation (rSO2) as a surrogate for pulsatile cerebral blood volume (CBV). In this study, the relationship between ICP- and rSO2-based indices of cerebrovascular reactivity, in a cohort of critically ill TBI patients, is explored using classical machine learning clustering techniques and multivariate time-series analysis.MethodsHigh-resolution physiologic data were collected in a cohort of adult moderate to severe TBI patients at a single quaternary care site. From this data both ICP- and rSO2-based indices of cerebrovascular reactivity were derived. Utilizing agglomerative hierarchical clustering and principal component analysis, the relationship between these indices in higher dimensional physiologic space was examined. Additionally, using vector autoregressive modeling, the response of change in ICP and rSO2 (ΔICP and ΔrSO2, respectively) to an impulse in change in arterial blood pressure (ΔABP) was also examined for similarities.ResultsA total of 83 patients with 428,775 min of unique and complete physiologic data were obtained. Through agglomerative hierarchical clustering and principal component analysis, there was higher order clustering between rSO2- and ICP-based indices, separate from other physiologic parameters. Additionally, modeled responses of ΔICP and ΔrSO2 to impulses in ΔABP were similar, indicating that ΔrSO2 may be a valid surrogate for pulsatile CBV.ConclusionsrSO2- and ICP-based indices of cerebrovascular reactivity relate to one another in higher dimensional physiologic space. ΔICP and ΔrSO2 behave similar in modeled responses to impulses in ΔABP. This work strengthens the body of evidence supporting the similarities between ICP-based and rSO2-based indices of cerebrovascular reactivity and opens the door to cerebrovascular reactivity monitoring in settings where invasive ICP monitoring is not feasible.

Relating clustered noise data to hospital patient satisfaction.

Hospital noise can be problematic for both patients and staff and consistently is rated poorly on national patient satisfaction surveys. A surge of research in the last two decades highlights the challenges of healthcare acoustic environments. However, existing research commonly relies on conventional noise metrics such as equivalent sound pressure level, which may be insufficient to fully characterize the fluctuating and complex nature of the hospital acoustic environments experienced by occupants. In this study, unsupervised machine learning clustering techniques were used to extract patterns of activity in noise and the relationship to patient perception. Specifically, nine patient rooms in three adult inpatient hospital units were acoustically measured for 24 h and unsupervised machine learning clustering techniques were applied to provide a more detailed statistical analysis of the acoustic environment. Validation results of five different clustering models found two clusters, labeled active and non-active, using k-means. Additional insight from this analysis includes the ability to calculate how often a room is active or non-active during the measurement period. While conventional LAeq was not significantly related to patient perception, novel metrics calculated from clustered data were significant. Specifically, lower patient satisfaction was correlated with higher Active Sound Levels, higher Total Percent Active, and lower Percent Quiet at Night metrics. Overall, applying statistical clustering to the hospital acoustic environment offers new insights into how patterns of background noise over time are relevant to occupant perception.

Seismic identification of carbonate reservoir sweet spots using unsupervised machine learning: A case study from Brazil deep water Aptian pre-salt data

The pre-salt high-yield carbonate reservoirs of Brazil are responsible for about 75% of national oil production. Still, seventeen years since its discovery, detailed seismic characterization of these reservoirs remains challenging, with producing and nonproducing lithologies often exhibiting the same amplitude response. Those reservoirs are complex, composed of high productivity lacustrine carbonates (stromatolites and grainstones), mixed with non-producing mudstones, clays, and igneous intrusions. To address this heterogeneity, we evaluate unsupervised machine learning clustering techniques including Self Organizing Maps (SOM), Generative Topographic Maps (GTM), and K-means and apply them to a suite of instantaneous, geometric, interval velocity, and AVO seismic attributes. We use both static and dynamic reservoir data from 13 wells at the Mero field that penetrated lacustrine carbonates from Barra Velha formation to validate the accuracy of our unsupervised learning models. These wells are excellent blind tests as they were not used for any algorithm training. In this pioneer work, the first done using machine learning in the Mero field, we find that SOM, GTM, and K-means models provide clusters that can be linked to sweet spots associated with fracturing, karstification, and hydrothermal alteration. Additionally, clusters related to sill intrusions, platform facies and distal fine grained non-reservoirs were identified. The results can be extrapolated to the adjacent Central Libra Area (still in exploration phase), where we have identified the same clusters of the sweet spots reservoirs in Mero Field and, thus, they can be considered as future targets. For accuracy assessment, the same attribute combination SOM clustering was performed in another Petrobras’ pre-salt confidential dataset and the results were satisfactory. So, for exploration projects optimization purposes, the new proposed workflows can also be applied to other areas with geological similarity to the Mero Field.

Applying unsupervised machine learning clustering techniques to early childcare soundscapes

Early childhood is a critical time period for language, brain, cognitive, and social/emotional development. Out-of-home childcare is a normative, typical experience for millions of young children. Although Indoor Environmental Quality (IEQ) in K-12 settings has received recent, significant attention, the links between IEQ and children’s learning and development in early childcare settings is a less understood topic. This work focuses specifically on the sound aspect of IEQ in early childcare settings to better understand typical noise levels and occupant experience. Standard approaches to analyzing background noise will be presented alongside more detailed statistical analyses utilizing unsupervised machine learning clustering techniques. Noise data collected in three daycares will be presented using typical acoustic metrics and clustering techniques to better understand room activity conditions and support new metrics. Overall, this study can lead to a better understanding of daycare soundscapes and pave the way towards a better childcare for young children.

Applying unsupervised machine learning clustering techniques to hospital soundscapes

Noise in hospitals can be problematic for both patients and staff and is consistently rated poorly on national patient satisfaction surveys. Previous studies have linked negative outcomes of hospital noise to numerous patient and staff challenges, such as reduced sleep and disrupted communication. Existing articles and guidelines commonly use equivalent sound pressure level as a primary noise metric. Additional insights into typical sound levels experienced by occupants can be found through more detailed statistical analyses of sound, such as by applying unsupervised machine learning clustering techniques. In this talk, clustering techniques will be explored in an effort to provide a more detailed analysis of the soundscape and various patterns of room activity. Noise data collected in three adult, inpatient hospital units will be analyzed using clustering techniques and compared against patient satisfaction scores. This more thorough, statistical characterization of the hospital soundscape can lead to better understanding of patterns of noise conditions and resultant occupant perceptions.

Integration of machine learning using hydroacoustic techniques and sediment sampling to refine substrate description in the Western Cape, South Africa

A method to map bio-physical benthic habitats on the continental shelf of South Africa has been developed and is being refined. The goal is to produce a benthic habitat classification method that bridges the disciplines of marine geophysics and biological science, with relevance to all elements of the local substrate, using modern methods. Here, we produced ground-truthed seafloor characterisation maps for two study areas (Koeberg Harbour and Clifton) in the Western Cape, South Africa. Multibeam bathymetry and backscatter data were collected and processed using machine learning clustering techniques. The study area offshore of Clifton was used to test the recently developed k-means clustering algorithm, and Koeberg Harbour, which is 35 km to the north, was used to validate the algorithm because sediment samples, along with drop camera footage, were integrated to better refine the results. Drop-camera footage was classified using the Collaborative and Automated Tools for Analysis of Marine Imagery (CATAMI) substrata classification scheme and sediment grab samples were processed using a settling tube and formulae based on the Wentworth (1922) and Folk and Ward (1957) statistics. The resulting statistics were used to define the sediment categories that were input into the clustering algorithm. The algorithm results show the distribution of sediment within the respective study areas based on the combination of inputs. Our work uses a stepwise approach from unsupervised methods (previously discussed in Pillay et al., 2020), to geological and hydroacoustic verification using ground-truthed data (discussed here), to integrated benthic habitat map production. This work focuses on the second step using hydroacoustic data and ground-truthed geological and sedimentological substrate data to create substrate maps. In the final step further hydroacoustic and biological investigations are needed in order to merge biological and geological habitats, to create benthic habitat maps, along the South African coastline.

Benthic habitat mapping from a machine learning perspective on the Cape St Francis inner shelf, Eastern Cape, South Africa

We have developed an algorithm to map benthic habitats on the continental shelf of South Africa, integrating marine geophysics and biological science. Multibeam bathymetry, backscatter and Remotely Operated Vehicle (ROV) footage were collected on the inner continental shelf of Cape St Francis and the hydroacoustic data were processed using machine learning clustering techniques. The k-means clustering algorithm was used to map the distribution of sediment at different depths. ROV footage was classified using the Collaborative and Automated Tools for Analysis of Marine Imagery (CATAMI) substrata classification scheme. Eight ROV dives along the three transects located off Seal Point, Cape St. Francis Point and within the bay were collected, and ranged from 30 to 80 m in depth. The most common Phyla in order from greatest to least abundance were; Cnidaria, Mollusca, Echinodermata, Chordata (fish species), Arthropoda (Subphylum Crustacea), Bryozoa, Porifera and Chordata (Class Ascidiacea), these were identified both on rocky substrate and sand. The first benthic habitat map of the Cape St. Francis area revealed ten different habitat types, accounting for geology, topography, and sediment cover. This work builds on an ongoing method development that incorporates broader habitat types from a range of substrates, enhancing the robustness of the algorithm, and will aid in improving our current understanding of the relationships between biota and physical habitats along the continental shelf of South Africa.

Phasor-based image segmentation: machine learning clustering techniques.

The phasor approach is a well-established method for data visualization and image analysis in spectral and lifetime fluorescence microscopy. Nevertheless, it is typically applied in a user-dependent manner by manually selecting regions of interest on the phasor space to find distinct regions in the fluorescence images. In this paper we present our work on using machine learning clustering techniques to establish an unsupervised and automatic method that can be used for identifying populations of fluorescent species in spectral and lifetime imaging. We demonstrate our method using both synthetic data, created by sampling photon arrival times and plotting the distributions on the phasor plot, and real live cells samples, by staining cellular organelles with a selection of commercial probes.

Automatic Image Segmentation Based on Phasor Space Clustering

We present our latest developments in automatic image segmentation for measurements on spectral and lifetime fluorescence imaging. The algorithm operates on the basis of the phasor approach and makes use of a well-established machine learning method for clustering N-dimensional data points, in our case the pixel data points on the phasor space. The technique is based on Gaussian mixture models solved by the use of the expectation-maximization algorithm. We show the method is the best candidate out of the many machine learning clustering techniques available due to the nature of the phasor data and also due to the spreading of points in the phasor space for the cases of low pixel photon counts. In order to quantify the power of the method we performed simulations with synthetic phasor data and we tested the method by measuring its performance against a series of established clustering techniques. We also show results on lifetime and spectral data obtained from samples of live cells. The samples were prepared by staining with a series of commercially-available organelle dyes in order to target distinct structures in the cells. We show how by taking a single spectral and lifetime acquisition we resolve up to six populations in the phasor space corresponding to the six cellular staining probes. Furthermore, we use this method in combination with our other recent work, where we make use of higher harmonics of the phasor transform in order to solve for the fractions of known components and also for lifetimes of unknown components. This multi-harmonic multi-component method is highly sensitive to noise and heterogeneous distributions and here is where our clustering method simplifies the problem by preprocessing the data and fragmenting it into pixels containing distinct lifetime/spectral characteristics. (Supported by NIH P41-GM103540).

Clustering Algorithm-Based Network Planning for Advanced Metering Infrastructure in Smart Grid

Nowadays, legacy electrical distribution systems are migrating to a new modern electric grid with the capability of supporting different applications such as advanced metering infrastructure (AMI), distributed energy resources, and electric vehicles. Among these applications, AMI is playing an important role in delivering data from customers to power utilities, supporting reliable real-time monitoring, and remote operation of power quality data and voltage profile. The AMI consists of smart meters, data aggregation points (DAPs), the utility control center, and communication networks. Appropriate network planning plays an essential role in facilitating the exchange of data between consumers and power utilities as well as accommodating new smart grid applications and future growth. This work proposes an optimal placement of DAPs for AMI based on machine learning clustering techniques in residential grids. Network partitioning is introduced to create sub-networks and graph algorithms generate a deployment topology given optimization constraints. A new measurement metric called coverage density is considered to indicate neighborhood area networks (NAN) zones with the appropriate coverage. Three real scenarios of NAN are considered: urban, suburban, and rural. The proposed algorithm is evaluated and compared with conventional heuristic optimization methods with respect to average and maximum distance between smart meters and DAPs, coverage density, and execution time.

Optimizing sizing schedules of aircrew helmets using machine learning techniques

Introduction: Aircrew equipment assemblies are life critical equipment worn by the aircrew to ensure protection from various life-threatening environmental conditions that may occur in nominal or off-nominal flight conditions. It is important for the sizing system of body worn aircrew equipment to be divided into different size classes such that it is as representative of existing groups in the population as feasible. Machine learning techniques provide a powerful tool to develop efficient sizing systems that are representative of actual body types existing in the population. The objective of the study was to use machine learning clustering techniques to identify naturally existing body types to formulate an optimal sizing system for an aircrew helmet. Material and Methods: The standard sizing schedule of an aircrew helmet using head length and head breadth was studied. An iterative semi-supervised machine learning protocol called K-means clustering was used to identify naturally occurring clusters of head types within the population. The cluster boundaries were identified to develop the final sizing schedule. Analysis of variance (ANOVA) with post hoc analysis was carried out on the four head lengths and head breadths. Results: Four clusters of head type were identified using K-means clustering. This led to a sizing schedule which can be descriptively called “Short-Narrow,” “Short-Broad,” “Long-Narrow,” “Long-Broad.” ANOVA showed a statistically significant difference between the four clusters for both head length and head breadth. Conclusion: Consideration of several variables in sizing schedules often fails to meet desired fitment. Standard sizing methods use linear partitions on individual control parameters. In reality, the proportional variation between parameters is not linear. Machine learning tools help in identification of naturally occurring clusters within the database considering several variables at the same time. This study uses machine learning techniques to identify existing groups in a population and develop a final sizing schedule. Similar analysis for other aircrew equipment assemblies can be carried out to determine sizing schedules which assist in design and fitment.

A Machine Learning Solution for Data Center Thermal Characteristics Analysis

The energy efficiency of Data Center (DC) operations heavily relies on a DC ambient temperature as well as its IT and cooling systems performance. A reliable and efficient cooling system is necessary to produce a persistent flow of cold air to cool servers that are subjected to constantly increasing computational load due to the advent of smart cloud-based applications. Consequently, the increased demand for computing power will inadvertently increase server waste heat creation in data centers. To improve a DC thermal profile which could undeniably influence energy efficiency and reliability of IT equipment, it is imperative to explore the thermal characteristics analysis of an IT room. This work encompasses the employment of an unsupervised machine learning technique for uncovering weaknesses of a DC cooling system based on real DC monitoring thermal data. The findings of the analysis result in the identification of areas for thermal management and cooling improvement that further feeds into DC recommendations. With the aim to identify overheated zones in a DC IT room and corresponding servers, we applied analyzed thermal characteristics of the IT room. Experimental dataset includes measurements of ambient air temperature in the hot aisle of the IT room in ENEA Portici research center hosting the CRESCO6 computing cluster. We use machine learning clustering techniques to identify overheated locations and categorize computing nodes based on surrounding air temperature ranges abstracted from the data. This work employs the principles and approaches replicable for the analysis of thermal characteristics of any DC, thereby fostering transferability. This paper demonstrates how best practices and guidelines could be applied for thermal analysis and profiling of a commercial DC based on real thermal monitoring data.

Short-range order structure motifs learned from an atomistic model of a Zr50Cu45Al5 metallic glass

The structural motifs of a Zr50Cu45Al5 metallic glass were learned from atomistic models using a new structure analysis method called motif extraction that employs point-pattern matching and machine learning clustering techniques. The motifs are the nearest-neighbor building blocks of the glass and reveal a well-defined hierarchy of structures as a function of coordination number. Some of the motifs are icosahedral or quasi-icosahedral in structure, while others take on the structure of the most close-packed geometries for each coordination number. These results set the stage for developing clearer structure-property connections in metallic glasses. Motif extraction can be applied to any disordered material to identify its structural motifs without the need for human input.

Reactivity of Amorphous Carbon Surfaces: Rationalizing the Role of Structural Motifs in Functionalization Using Machine Learning.

Systematic atomistic studies of surface reactivity for amorphous materials have not been possible in the past because of the complexity of these materials and the lack of the computer power necessary to draw representative statistics. With the emergence and popularization of machine learning (ML) approaches in materials science, systematic (and accurate) studies of the surface chemistry of disordered materials are now coming within reach. In this paper, we show how the reactivity of amorphous carbon (a-C) surfaces can be systematically quantified and understood by a combination of ML interatomic potentials, ML clustering techniques, and density functional theory calculations. This methodology allows us to process large amounts of atomic data to classify carbon atomic motifs on the basis of their geometry and quantify their reactivity toward hydrogen- and oxygen-containing functionalities. For instance, we identify subdivisions of sp and sp2 motifs with markedly different reactivities. We therefore draw a comprehensive, both qualitative and quantitative, picture of the surface chemistry of a-C and its reactivity toward −H, −O, −OH, and −COOH. While this paper focuses on a-C surfaces, the presented methodology opens up a new systematic and general way to study the surface chemistry of amorphous and disordered materials.

Measuring the Feasibility of Clustering Techniques on Usability Performance Data

This paper proposes a methodology that utilizes unsupervised machine learning clustering techniques in performance based usability data. This paper will first discuss the introduction and current works in the aforementioned domain followed by proposing the methodology to compare and find a better clustering algorithm in processing usability performance data in the field of mobile augmented reality interfaces. The paper will then present the results yield from an experiment abiding by the proposed methodology and discusses the analysis. The paper will end with a short discussion followed by proposed future works in the research area. Keywords: Augmented Reality, Clustering, Machine Learning, Mobile, Usability

The Analysis of Anticancer Drug Sensitivity of Lung Cancer Cell Lines by using Machine Learning Clustering Techniques

Lung cancer is the commonest type of cancer with the highest fatality rate worldwide. There is continued research that experiments on drug development for lung cancer patients by assessing their responses to chemotherapeutic treatments to select novel targets for improved therapies. This study aims to analyze the anticancer drug sensitivity in human lung cancer cell lines by using machine learning techniques. The data for this analysis is extracted from the National Cancer Institute (NCI). This experiment uses 408,291 human small molecule lung cancer cell lines to conclude. The values are drawn from describing the raw viability values for 91 human lung cancer cell lines treated with 354 different chemical compounds and 432 concentration points tested in each replicate experiments. Our analysis demonstrated the data from a considerable amount of cell lines clustered by using Simple K-means, Filtered clustering and by calculating sensitive drugs for each lung cancer cell line. Additionally, our analysis also demonstrated that the Neopeltolide, Parbendazole, Phloretin and Piperlongumine anti-drug chemical compounds were more sensitive for all 91 cell lines under different concentrations (p-value < 0.001). Our findings indicated that Simple K-means and Filtered clustering methods are completely similar to each other. The available literature on lung cancer cell line data observed a significant relationship between lung cancer and anticancer drugs. Our analysis of the reported experimental results demonstrated that some compounds are more sensitive than other compounds; Phloretin was the most sensitive compound for all lung cancer cell lines which were nearly about 59% out of 91 cell lines. Hence, our observation provides the methodology on how anticancer drug sensitivity of lung cancer cell lines can be analyzed by using machine learning techniques, such as clustering algorithms. This inquiry is a useful reference for researchers who are experimenting on drug developments for the lung cancer in the future.

Automatic characterization of ignition processes with machine learning clustering techniques

AbstractMachine learning clustering techniques are used to characterize and, after the training phase, to identify phases within an ignition process. For the ethanol mechanism used in this paper, four physically identifiable phases were found and characterized: the initiation phase, preignition phase, ignition phase, and the postignition phase. The clustering is done with respect to fuzzy logic predicates identifying the maxima, minima, and inflection points of the species profiles. The cluster descriptions characterize the phases found and are in human interpretable form. In addition, these descriptions are powerful enough to be used to predict the phase structure under new conditions. Cluster phases were calculated for the ethanol mechanism at an equivalence ratio of 0.5, a pressure of 3.3 bar, and the temperatures 1200, 1300, 1400, and 1500 K. The resulting cluster phase descriptions were then successfully used to predict the phase structure and ignition delay times for other temperatures in the range from 1200 to 1500 K. The effect of different fuzzy logic predicate profile descriptions is studied to emphasize that the boundaries of some phases, specifically that between the preignition and the ignition phase, are a matter of what the modeler considers important. The end of the ignition phase corresponds to the ignition delay time and was relatively independent of the predicate descriptions used to determine the phases. © 2006 Wiley Periodicals, Inc. Int J Chem Kinet 38: 621–633, 2006

Characterizing complex reaction mechanisms using machine learning clustering techniques

AbstractA machine learning conceptual clustering method applied to reaction mechanisms provides an automatic and, hence, unbiased means to differentiate between reactive phases within a total reactive process. Similar reactive phases were defined by means of local reaction sensitivity values. The method was applied to the Hochgreb and Dryer aldehyde combustion mechanism of 36 reactions. Three major time ranges were found and characterized: an initial phase of aldehyde reaction, an intermediate phase where only a small amount of aldehyde is left, and an end phase of reactions to final products. Further refinements of these phases into subtime intervals were found. All ranges found could be chemically justified. This method is meant as a supplement to existing methods of mechanism analysis and its main purpose is the automatic characterization of existing mechanisms and can potentially be used for mechanism reduction. © 2003 Wiley Periodicals, Inc. Int J Chem Kinet 36: 107–118, 2004