Driven Machine Learning Research Articles

Comparison of the diagnostic accuracy of resting-state fMRI driven machine learning algorithms in the detection of mild cognitive impairment

Mild cognitive impairment (MCI) is a potential therapeutic window in the prevention of dementia; however, automated detection of early cognitive deterioration is an unresolved issue. The aim of our study was to compare various classification approaches to differentiate MCI patients from healthy controls, based on rs-fMRI data, using machine learning (ML) algorithms. Own dataset (from two centers) and ADNI database were used during the analysis. Three fMRI parameters were applied in five feature selection algorithms: local correlation, intrinsic connectivity, and fractional amplitude of low frequency fluctuations. Support vector machine (SVM) and random forest (RF) methods were applied for classification. We achieved a relatively wide range of 78–87% accuracy for the various feature selection methods with SVM combining the three rs-fMRI parameters. In the ADNI datasets case we can also see even 90% accuracy scores. RF provided a more harmonized result among the feature selection algorithms in both datasets with 80–84% accuracy for our local and 74–82% for the ADNI database. Despite some lower performance metrics of some algorithms, most of the results were positive and could be seen in two unrelated datasets which increase the validity of our methods. Our results highlight the potential of ML-based fMRI applications for automated diagnostic techniques to recognize MCI patients.

Blood Count Scattergrams Are Fingerprints of Blood: Using AI to Inform Health Status

Introduction: Blood count analysers are routinely used to assess health status and for disease monitoring. The major analysers generate more than 300 red cell, leukocyte and platelet parameters. However, in clinical practice, only 10-20 are utilised; a greater number are reviewed by laboratories but some are merely for ensuring instrument stability and quality assurance. Although there are hundreds more measurable parameters available through extended analytical channels, these are not assessed due to the complexity of the numerical analysis and questions as to the clinical utility. However, these extended analytical parameters have potential to be used to generate real-time ‘3-Dimensional (3-D)‘ details of blood cells. Aim: In this study we assessed 312 parameters generated from 5,800 patient samples (of all ages and genders) at a tertiary care hospital (National Institute of Blood Disease, Karachi-Pakistan). These were anonymously outputted and processed by machine learning (ML). Method: The methodology of the present study was based on the waterfall model. The output data from a haematology analyser (Sysmex XN-1000, Kobe Japan) in CSV format, having a total 433 columns, was pre-processed to remove unnecessary features (such as date, analysis details, rack position, receiving time, alerts and others) using Pandas and Numpy libraries, and where required scaling was performed. Data labeling was conducted on conclusions reported on their respective confirmatory tests. The processed data (which included 312 features of the total 433 columns) was labeled (with 67 conclusions on their confirmatory tests). The extracted data was fed to the artificial intelligence Machine Learning models through Python modules and libraries for training, testing, and validation purposes. The web application was developed on modern Python framework (Flask) to automate and provide an option of ‘drag and drop’ the CSV file exported from analyser, we connected pre-processing (data engineering), Machine Learning, and Prediction view by set of different tools (majorly JavaScript libraries). This generated various metrics including accuracy, precision, recall and the mean of precision and recall (F1 score) in relation to the prediction (results) from the CSV file submitted to our system. For each forecast the entire precision report and prediction box were added an optional visualisation on our web panel. Results: Analysis of 1.8 million data points (312 parameters x 5,800 samples) presented promising predictive potential, as, on principal component analysis (PCA) pilot the total variance was remained 41.6% showing that a linear combination of parameters can explain much variability. On a heat map the clustering and visualisation advocated the predictive potential and signatory deviational trends (fingerprints) respectively of these 3-D blood cell features. Examples included separation of myeloid from lymphoid, chronic from acute, bacteria from viral, deficiency of iron from deficiency from vitamin B12 / Folic acid, and differentiation of haemoglobinopathies. The patterns of normal, immature and abnormal blood cells under the title of cell population data was well demonstrated from results of our machine learning models. Of note, we observed an accuracy of 85.6% along with 91.2% precision for one of the ML models used (Random Forest Classifier). Conclusion: The opportunities and challenges of such high dimensional cell population data derived from a complete blood count can provide a novel patient-specific haematological fingerprint. This extended deviational patterning (fingerprint) can provide interpretive diagnostic data with practical disease-specific patterns. This pilot study shows that complete blood count data driven machine learning applications has great potential to uncover disease-associated patterns which could be applied in practice. It also has capacity to provide baseline testing would could assist in sequential health monitoring and potentially the generation of personal reference ranges.

Predicting real-time within-vehicle air pollution exposure with mass-balance and machine learning approaches using on-road and air quality data

Modelling the air pollutant concentrations within-vehicles is an essential step to estimate our daily exposure to air pollution. This is a challenging issue however, since the processes that affect the exposures within-vehicles change with different driving patterns and ventilation settings. This study introduces an innovative approach that combines mass-balance principles and machine learning techniques, leveraging ambient air quality, on-road and within-vehicle measurements of particulate matter (PM10, PM2.5, PM1), nitrogen dioxide (NO2), nitrogen oxides (NOx), aerosol lung surface deposited area (LSDA) and ultrafine particles (UFP) under different ventilation settings to estimate air pollution exposure levels within vehicles. The first model (MB) includes basic physical and chemical processes and follows a mass-balance approach to estimate the within-vehicle concentrations. The second model (ML) applies data driven machine learning algorithms to a training set of observations to predict unseen within-vehicle concentrations. By using a number generator, the whole observational dataset was divided to 80:20 and 80% was used to build and train the ML model, while 20% was used for validation. Both models demonstrated good predictions of observations apart from an underestimation in UFP and LSDA. The ML model showed better predictive power than the MB model and had skill in predicting the unseen within-vehicle exposures. The ML model predictions were as good as the MB model for most of the species and improved for NO2. The ML model demonstrated good index of agreement (IOA >0.69) and Pearson correlation coefficient (r > 0.80) for all the species. The inclusion of air quality data from nearby monitoring stations instead of on-road (sampled while driving), in the ML model showed promising and new capabilities to within-vehicle exposure predictions. In an era where air pollution is a growing concern, understanding and predicting within-vehicle air pollution exposure is of great importance for public health and environmental research. This research not only advances the field of exposure assessment but (at no extra cost) also demonstrates practical implications for real-time exposure mapping and health impact assessment of vehicle occupants with existing infrastructure.

Deep Learning–Based Summertime Turbulence Intensity Estimation Using Satellite Observations

Abstract Turbulence is what we want to avoid the most during flight. Numerical weather prediction (NWP) model–based methods for diagnosing turbulence have offered valuable guidance for pilots. NWP-based turbulence diagnostics show high accuracy in detecting turbulence in general. However, there is still room for improvements such as capturing convectively induced turbulence. In such cases, observation data can be beneficial to correctly locate convective regions and help provide corresponding turbulence information. Geostationary satellite data are commonly used for upper-level turbulence detection by utilizing its water vapor band information. The Geostationary Operational Environmental Satellite (GOES)-16 carries the Advanced Baseline Imager (ABI), which enables us to observe further down into the atmosphere with improved spatial, temporal, and spectral resolutions. Its three water vapor bands allow us to observe different vertical parts of the atmosphere, and from its infrared window bands, convective activity can be inferred. Such multispectral information from ABI can be helpful in inferring turbulence intensity at different vertical levels. This study develops U-Net based machine learning models that take ABI imagery as inputs to estimate turbulence intensity at three vertical levels: 10–18, 18–24, and above 24 kft (1 kft ≈ 300 m). Among six different U-Net-based models, U-Net3+ model with a filter size of three showed the best performance against the pilot report (PIREP). Two case studies are presented to show the strengths and weaknesses of the U-Net3+ model. The results tend to be overestimated above 24 kft, but estimates of 10–18 and 18–24 kft agree well with the PIREP, especially near convective regions. Significance Statement Turbulence is directly related to aviation safety as well as cost-effective aircraft operation. To avoid turbulence, turbulence diagnostics are calculated from numerical weather prediction (NWP) model outputs and are provided to pilots. The goal of this study is to develop a satellite data–driven machine learning model that estimates turbulence intensity in three different vertical layers to provide additional information along with the NWP-based turbulence diagnostics. Validation results against pilot reports show that the machine learning model performs comparable to NWP-based turbulence diagnostics. Furthermore, results with different channel selections reveal that using multiple water vapor channels can help extract additional information for estimating turbulence intensity at lower levels.

Model and Data Driven Machine Learning Approach for Analyzing the Vulnerability to Cascading Outages With Random Initial States in Power Systems

In this paper, a hybrid machine learning model is applied to evaluate the relationship between random initial states and the power system’s vulnerability to cascading outages. A cascading outage simulator (CS), which uses off-line AC power flows, is proposed for generating training data. The initial states are randomly selected and the CS model is deployed for each initial state, where power system generation and loads are adjusted dynamically and power flows are redistributed to quantify the vulnerability metric. Furthermore, the proposed hybrid machine learning model deploys a combined Support Vector Machine (SVM) classification and Gradient Boosting Regression (GBR) to improve the learning precision. The classification model is trained by SVM, which divides the data into two categories with and without load shedding. Then, GBR is adopted only for the data with load shedding to determine the relationship between input power outage states and the vulnerability metric. The proposed vulnerability analysis approach is applied to several test systems and the results are analyzed. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —The power system vulnerability can be quantified by cascading outage simulations. However, there are two challenges: i) there are a huge number of possible initial states and we cannot enumerate all these initial states for the cascading outage simulation. Neither can we precisely quantify the bus vulnerability. ii) The cascading outage simulation may be time-consuming for large-scale power systems, which is challenging for the online application. To address the above challenges, we expect to design a machine learning technique to predict the power system vulnerability, which can train the model in an offline way and then use it for the online application. Firstly, since there is not enough operation data from practical power systems, we develop a cascading outage simulator, using off-line AC power flows, for generating synthetic training data. Secondly, we observe that the training precision by directly applying the regression model may be very poor because the output of the machine learning model may take on an uneven distribution concerning input parameters. Thus, we propose a hybrid machine learning model with a combined classification and regression method, where the classification model is employed to remove the data without the load shedding, and the regression model then determines the relationship between input power outage states and the vulnerability metric. The proposed model and method have been tested on several systems including a practical large-scale Polish power system to show the effectiveness.

Low-cycle fatigue life assessment of SAC solder alloy through a FEM-data driven machine learning approach

Purpose This paper aims to present the novel stacked machine learning approach (SMLA) to estimate low-cycle fatigue (LCF) life of SAC305 solder across structural parts. Using the finite element simulation (FEM) and continuous damage mechanics (CDM) model, a fatigue life database is built. The stacked machine learning (ML) model's iterative optimization during training enables precise fatigue predictions (2.41% root mean square error [RMSE], R2 = 0.975) for diverse structural components. Outliers are found in regression analysis, indicating potential overestimation for thickness transition specimens with extended lifetimes and underestimation for open-hole specimens. Correlations between fatigue life, stress factors, nominal stress and temperature are unveiled, enriching comprehension of LCF, thus enhancing solder behavior predictions. Design/methodology/approach This paper introduces stacked ML as a novel approach for estimating LCF life of SAC305 solder in various structural parts. It builds a fatigue life database using FEM and CDM model. The stacked ML model iteratively optimizes its structure, yielding accurate fatigue predictions (2.41% RMSE, R2 = 0.975). Outliers are observed: overestimation for thickness transition specimens and underestimation for open-hole ones. Correlations between fatigue life, stress factors, nominal stress and temperature enhance predictions, deepening understanding of solder behavior. Findings The findings of this paper highlight the successful application of the SMLA in accurately estimating the LCF life of SAC305 solder across diverse structural components. The stacked ML model, trained iteratively, demonstrates its effectiveness by producing precise fatigue lifetime predictions with a RMSE of 2.41% and an “R2” value of 0.975. The study also identifies distinct outlier behaviors associated with different structural parts: overestimations for thickness transition specimens with extended fatigue lifetimes and underestimations for open-hole specimens. The research further establishes correlations between fatigue life, stress concentration factors, nominal stress and temperature, enriching the understanding of solder behavior prediction. Originality/value The authors confirm the originality of this paper.

An application of ontology driven machine learning model challenges for the classification of social media data: a systematic literature review

This systematic literature review aimed to explore the challenges and limitations of applying ontology driven machine learning models to the classification of social media data. Social media platforms generate a vast amount of data that requires automated and reliable classification to facilitate analysis and decision-making. Ontology driven machine learning models offer a promising approach to address this need by harnessing the power of both ontologies and machine learning algorithms to improve accuracy and efficiency. However, the application of such models to social media data classification poses unique challenges due to the complex and dynamic nature of social media data. To address this research gap, a systematic literature search was conducted, and 20 studies were included in the review. The findings of this review suggest that ontology driven machine learning models offer a promising approach to address the challenges of social media data classification. However, the existing literature highlights several challenges that need to be addressed, such as ontology development, feature selection, and model validation. Overall, the review provides insights into the current state of research on ontology driven machine learning models for social media data classification, identifies research gaps, and suggests directions for future investigation.

Three-step learning strategy for designing 15Cr ferritic steels with enhanced strength and plasticity at elevated temperature

15Cr ferrite steels are urgently required in advanced Ultra-supercritical power plants but meet design challenges in balancing excellent strength and plasticity at high temperatures. We developed a three-step learning strategy based on mutually driven machine learning and purposeful experiments to complete this multi-objective task. Compared with traditional adaptive learning and local-interpolation learning, this step-by-step modular manner provides good transparency and interpretability of the information flow, which is ensured by identifying essential factors from an exquisitely prepared composition-microstructure dataset, and learning valuable knowledge about the composition-property relationship. The requirement of only two groups of experiments indicates the low cost and high efficiency of the strategy. Performing the strategy, we found that Ti is another key element affecting the Laves phase besides Mo and W, and their effects on ultimate tensile strength (UTS) and elongation were also uncovered. Importantly, several low-cost steels free of Co were successfully designed, and the best steel exhibited 156%, 31%, and 62% higher UTS and elongation at 650 °C than the typical 9Cr, 15Cr, and 20Cr steels, respectively. Based on the advantages and success of the strategy in terms of alloy improvement, we believe the strategy suits other multi-objective design tasks in more materials systems.

CAMERA-BASED REMOTE PHOTOPLETHYSMOGRAPHY TO PREDICT BLOOD PRESSURE IN CLINIC PATIENTS WITH CARDIOVASCULAR DISEASE

Objective: Remote photoplethysmography (rPPG) uses a camera to capture peripheral blood flow. It remains to be seen if rPPG can accurately predict blood pressure (BP). Most studies have included healthy, normotensive subjects. This is the first investigation of rPPG to predict BP for patients with cardiovascular disease (CVD). Design and method: 132 patient visits were completed in a cardiology clinic between March and December, 2022 (Table 1). Patients were seated while a digital camera recorded two 2.5 minute videos of the face and palm. BP was measured by an oscillometric cuff from the left arm at the beginning, midpoint, and end of the session. Continuous heart rate and rhythm were measured by a single-lead ECG from electrodes on the right and left wrists. Continuous PPG was measured by a contact sensor on the left index finger. Videos were cropped and spatially-averaged to extract the green channel, which serves as a proxy for the rPPG signal. Pulse segmentation and selection were used to identify five high-quality consecutive beats. Raw rPPG signal was analyzed with its first and second derivatives into a one-dimensional convolutional neural network. The network is trained to predict systolic and diastolic BP by minimizing the mean squared error between the predicted video-based contactless BP and cuff-based BP measurement. A 5-fold subject independent cross-validation was performed. Results: There is modest correlation between rPPG-predicted and measured systolic BP (r = 0.47, mean error -0.47 ± 15.5 mm Hg), but weaker correlation for diastolic BP (r = 0.16, mean error -0.80 ± 8.1 mm Hg). Conclusions: rPPG shows potential for BP prediction in real-world patients with CVD. Future work will leverage larger data driven machine learning methods to improve the rPPG signal quality and examine the impact on BP measurements. Robust training data, including extreme values of blood pressure and a variety of patient demographics, are needed to improve these models. Major limitations include lack of guideline-directed protocols for rPPG validation for BP prediction and inability to assess temporal correlation between rPPG signal and BP.

Design of high-performance high-entropy nitride ceramics via machine learning-driven strategy

High-entropy nitride (HEN) ceramics have exhibited excellent mechanical properties in comparison to transition metal nitrides. However, the huge unexplored compositional space makes the trial-and-error experiments and first-principles calculations cost highly. In this work, we proposed a data augmentation generative adversarial network (DAGAN)-driven machine learning (ML) design strategy to predict the hardness, modulus and wear resistance of novel HEN compositions. Several feature selection algorithms were compared to select the optimal descriptor combination. To overcome the data shortage problem of HENs, we established a property-conditioned DAGAN and the accuracies of ML models were maximumly increased by up to 14.67%. Eight super-hard HEN systems with the hardness above 40 GPa were found among the compositional space, in which seven have yet to be experimentally synthesized. The intrinsic effects of chemical descriptors were further explored through ternary property diagrams, which provides an efficient guidance for the design of novel high-performance HENs.

Machine Learning Based Risk Prediction for Major Adverse Cardiovascular Events for ELGA-Authorized Clinics1.

Artificial Intelligence (AI) has had an important impact on many industries as well as the field of medical diagnostics. In healthcare, AI techniques such as case-based reasoning and data driven machine learning (ML) algorithms have been used to support decision-making processes for complex tasks. This is used to assist medical professionals in making clinical decisions. A way of supporting clinicians is providing predicted prognoses of various ML models. Training an ML model based on the data of a hospital and using it on another hospital have some challenges. In this research, we applied data analysis to discover required data filters on a hospital's EHR data for training a model for another hospital. We applied experiments on real-world data of ELGA (Austrian health record system) and KAGes (a public healthcare provider of 20+ hospitals in Austria). In this scenario, we train the prediction model for ELGA- authorized health service providers using the KAGes data since we do not have access to the complete ELGA data. Finally, we observed that filtering the data with both feature and value selection increases the classification performance of the prediction model, which is trained for another system.

Data driven machine learning models for short‐term load forecasting considering electrical vehicle load

AbstractElectric vehicles (EVs) are gaining popularity due to their fuel efficiency and ability to reduce greenhouse gas emissions. Significant penetration of EVs with unregulated charging, which can have a substantial impact on power networks. Accurate load predictions, including the charging of EVs are crucial for ensuring the cost‐effective and dependable operation of power systems. In order to estimate short‐term load in the presence of EV load, the XG (extreme gradient) boost algorithm is proposed and the effectiveness of the performances are checked against other models. A variety of distinct meteorological parameters and the electrical load pattern for the years 2017 and 2018 for Northeast India are used to train the machine learning classifier models. R‐squared value analyses are also performed to identify the most correlated input parameters that influence the results of various models. Analysis shows that temperature, cloud cover, heat index, dew point, wind chill, and perceived temperature are substantially connected with electricity consumption. The performance of XG boost is outperformed by comparing prediction results with decision tree, random forest, K nearest neighbors, and logistic regression. Three separate case studies were employed for verification using the precision, F1 score, sensitivity, specificity, and accuracy metrics. Our findings demonstrate that, in comparison to other forecasting models, XG Boost exhibits higher accuracy (83.84%, 81.51%, and 85.97%) and robust forecasts.

Using machine learning to determine the time of exposure to infection by a respiratory pathogen

Given an infected host, estimating the time that has elapsed since initial exposure to the pathogen is an important problem in public health. In this paper we use longitudinal gene expression data from human challenge studies of viral respiratory illnesses for building predictive models to estimate the time elapsed since onset of respiratory infection. We apply sparsity driven machine learning to this time-stamped gene expression data to model the time of exposure by a pathogen and subsequent infection accompanied by the onset of the host immune response. These predictive models exploit the fact that the host gene expression profile evolves in time and its characteristic temporal signature can be effectively modeled using a small number of features. Predicting the time of exposure to infection to be in first 48 h after exposure produces BSR in the range of 80–90% on sequestered test data. A variety of machine learning experiments provide evidence that models developed on one virus can be used to predict exposure time for other viruses, e.g., H1N1, H3N2, and HRV. The interferon alpha /beta signaling pathway appears to play a central role in keeping time from onset of infection. Successful prediction of the time of exposure to a pathogen has potential ramifications for patient treatment and contact tracing.

Prototype Theory Meets Word Embedding: A Novel Approach for Text Categorization via Granular Computing

The problem of the information representation and interpretation coming from senses by the brain has plagued scientists for decades. The same problems, from a different perspective, hold in automated Pattern Recognition systems. Specifically, in solving various NLP tasks, an ever better and richer semantic representation of text as a set of features is needed and a plethora of text embedding techniques in algebraic spaces are continuously provided by researchers. These spaces are well suited to be conceived as conceptual spaces in light of the Gärdenfors’s Conceptual Space theory, which, within the Cognitive Science paradigm, seeks a geometrization of thought that bridges the gap between an associative lower level and a symbolic higher level in which information is organized and processed and where inductive reasoning is appropriate. Granular Computing can offer the toolbox for granulating text that can be represented by more abstract entities than words, offering a good hierarchical representation of the text embedded in an algebraic space driving Machine Learning applications, specifically, in text mining tasks. In this paper, the Conceptual Space Theory, the Granular Computing approach and Machine Learning are bound in a novel common framework for solving some text categorization tasks with both standard classifiers suited for working with mathbb {R}^n vectors and a Recurrent Neural Network (RNN) — an LSTM — able to deal with sequences. Instead of working with word vectors, the algorithms process more abstract entities (concepts), where patterns, in a first approach, are obtained through the construction of a symbolic histogram starting from a suitable set of information granules, representing a document as a distribution of concepts. For the RNN case, as a further novelty, a text is represented as a random walk over prototypes within the conceptual space synthesized over a suitable text embedding procedure. A comparison of the performance and a critical discussion are offered for both a neural embedding technique and the well-known LSA, showing how the conceptual level leads also to Knowledge Discovery applications.

Prediction of dose deposition matrix using voxel features driven machine learning approach.

A dose deposition matrix (DDM) prediction method using several voxel features and a machine learning (ML) approach is proposed for plan optimization in radiation therapy. Head and lung cases with the inhomogeneous medium are used as training and testing data. The prediction model is a cascade forward backprop neural network where the input is the features of the voxel, including 1) voxel to body surface distance along the beamlet axis, 2) voxel to beamlet axis distance, 3) voxel density, 4) heterogeneity corrected voxel to body surface distance, 5) heterogeneity corrected voxel to beamlet axis, and (6) the dose of voxel obtained from the pencil beam (PB) algorithm. The output is the predicted voxel dose corresponding to a beamlet. The predicted DDM was used for plan optimization (ML method) and compared with the dose of MC-based plan optimization (MC method) and the dose of pencil beam-based plan optimization (PB method). The mean absolute error (MAE) value was calculated for full volume relative to the dose of the MC method to evaluate the overall dose performance of the final plan. For patient with head tumor, the ML method achieves MAE value 0.49 × 10-4 and PB has MAE 1.86 × 10-4. For patient with lung tumor, the ML method has MAE 1.42 × 10-4 and PB has MAE 3.72 × 10-4. The maximum percentage difference in PTV dose coverage (D98) between ML and MC methods is no more than 1.2% for patient with head tumor, while the difference is larger than 10% using the PB method. For patient with lung tumor, the maximum percentage difference in PTV dose coverage (D98) between ML and MC methods is no more than 2.1%, while the difference is larger than 16% using the PB method. In this work, a reliable DDM prediction method is established for plan optimization by applying several voxel features and the ML approach. The results show that the ML method based on voxel features can obtain plans comparable to the MC method and is better than the PB method in achieving accurate dose to the patient, which is helpful for rapid plan optimization and accurate dose calculation. Establishment of a new machine learning method based on the relationship between the voxel and beamlet features for dose deposition matrix prediction in radiation therapy.

Machine learning based fault detection and state of health estimation of proton exchange membrane fuel cells

In fuel cells, chemical energy is directly converted into heat and electricity without any emissions which makes them an attractive substitute for various energy needs. Fuel cells have high energy conversion ratio and high-power densities which make them suitable for automotive applications. However, these fuel cell systems suffer with low reliability and durability as system components develop faults during operation resulting in degradation and diminished system performance. In this context, fault detection and fault mitigation strategies are being extensively developed. Diagnostic approaches like electrochemical impedance spectroscopy, cyclic voltammetry, and galvanostatic analysis offer a truthful representation of the State of Health (SOH) of the fuel cell. However, these approaches are intrusive and require pausing the operation of the fuel cell effecting its integrity. Machine learning based fault detection and SOH estimation is a non-intrusive approach where a mapping function is established between the indicators and SOH. The SOH of a fuel cell can be correlated to the patterns in sensor signals or indicators. Indicators that influence SOH are cell voltages, current density distribution, impedance spectra, acoustic emission and magnetic fields. Developing an accurate fault detection and state estimation technique through data driven machine learning approaches will allow corrective measures to avoid irreversible faults and improve the reliability and durability of fuel cells.

MSR19 An Update on Real-Time Application of Machine Learning Programs to Improve Cardiovascular Risk Prediction in European Population

According to European Society of Cardiology, cardiovascular disease (CVD) accounts for 3.9 million deaths/year in Europe. In recent years, artificial intelligence (AI)-driven machine learning (ML) has shown promise in CVD risk prediction. This study aimed to summarise the composite predictive ability of AI/ML algorithms to improve CVD risk prediction in focused populations.

A data driven machine learning approach to differentiate between autism spectrum disorder and attention-deficit/hyperactivity disorder based on the best-practice diagnostic instruments for autism

Autism spectrum disorder (ASD) and attention-deficit/hyperactivity disorder (ADHD) are two frequently co-occurring neurodevelopmental conditions that share certain symptomatology, including social difficulties. This presents practitioners with challenging (differential) diagnostic considerations, particularly in clinically more complex cases with co-occurring ASD and ADHD. Therefore, the primary aim of the current study was to apply a data-driven machine learning approach (support vector machine) to determine whether and which items from the best-practice clinical instruments for diagnosing ASD (ADOS, ADI-R) would best differentiate between four groups of individuals referred to specialized ASD clinics (i.e., ASD, ADHD, ASD + ADHD, ND = no diagnosis). We found that a subset of five features from both ADOS (clinical observation) and ADI-R (parental interview) reliably differentiated between ASD groups (ASD & ASD + ADHD) and non-ASD groups (ADHD & ND), and these features corresponded to the social-communication but also restrictive and repetitive behavior domains. In conclusion, the results of the current study support the idea that detecting ASD in individuals with suspected signs of the diagnosis, including those with co-occurring ADHD, is possible with considerably fewer items relative to the original ADOS/2 and ADI-R algorithms (i.e., 92% item reduction) while preserving relatively high diagnostic accuracy. Clinical implications and study limitations are discussed.

Machine learning model to predict endophytic colonisation of rice cultivar plant tissues by Beauveria bassiana isolates and their potential as bio-control agents against rice stem borer using existing knowledge

Background: Finding well-known Beauveria bassiana isolates that could preserve rice crops from Sesamia calamistis (stem borer) is problematic. Another difficult task is the development of precise inoculation methods, which have been employed for their establishment as endophytes in cereal crops. This study proposed machine learning models to predict the best entomopathogenic fungi, Beauveria bassiana that could directly protect rice crops against Sesamia calamistis. Methods: Data driven machine learning decisions were implemented and assessed from 60 experimental runs with nine different feature/input variables and three target/output variables following foliar spray and seed treatment inoculation method. The feature variables consisted of rice plant tissue, such as Nerica-L19, Nerica1, Nerica8, the time, and the five promising isolates Beauveria bassiana (Bb3, Bb4, Bb10, Bb21, Bb35). The target variable consisted of the number of colonised roots, stems and leaves, expressed as a percentage depending on the degree of protection after each inoculation. A data driven decision by the extreme gradient boosting regression algorithm was used to proficiently abstract the situation where there is no direct relationship between features and target variables. Results: The foliar spray inoculation method exhibited high coefficient of determination (R2) of 0.99, 0.98 and 0.94 depending on the number of colonised stems, roots and leaves, respectively, while the seed treatment approach exhibited the coefficient of determination (R2) of 0.91, 0.87 and 0.75, respectively. Conclusions: These results demonstrated that the Extreme Gradient Boosting algorithm effectively abstracted the nonlinear relationship between the attribute variables that were taken into consideration and predicted Beauveria bassiana as a bio-pesticide for rice and perhaps other cereal stem borers. Thus, this XGBoost regression model could be used to navigate the optimization domain and reduce the development time of the biocontrol process.

Prediction of pullout interaction coefficient of geogrids by extreme gradient boosting model

Geogrids embedded in fill materials are checked against pullout failure through standard pullout testing methodology. The test determines the pullout interaction coefficient which is critical in fixing the embedment length of geogrids in mechanically stabilized earth walls. This paper proposes prediction of pullout interaction coefficient using data driven machine learning regression algorithms. The study primarily focusses on using extreme gradient boosting (XGBoost) method for prediction. A data set containing 220 test results from the literature has been used for training and testing. Predicted results of XGBoost have been compared with the results of random forest (RF) ensemble learning based algorithm. The predictions of XGBoost model indicates 85% accuracy and that of RF model shows 77% accuracy, indicating significantly superior and robust prediction through XGBoost above RF model. The importance analysis indicates that normal stress is the most significant factor that influences the pullout interaction coefficients. Subsequently pullout tests have been performed on geogrid embedded in four different fill materials at three normal stresses. The proposed XGBoost model gives 90% accuracy in prediction of pullout interaction coefficient compared to laboratory test results. Finally, an open-source graphical user interface based on the XGBoost model has been created for preliminary estimation of the pullout interaction coefficient of geogrid at different test conditions.