Tree Machine Learning Algorithm Research Articles

Machine Learning Distinguishes Familiar from Unfamiliar Pairs of Subjects Performing an Eye Contact Task: A Systemic Physiology Augmented Functional Near-Infrared Spectroscopy Hyperscanning Study.

Eye contact is an important aspect of human communication and social interactions. Changes in brain and systemic physiological activity associated with interactions between humans can be measured with systemic physiology augmented functional near-infrared spectroscopy (SPA-fNIRS) hyperscanning, enabling inter-brain and inter-body synchronisation to be determined. In a previous study, we found that pairs of subjects that are socially connected show higher brain and body synchrony. To enable a deeper understanding, our aim was to build and automatically detect the best set of features to distinguish between two different groups (familiar and unfamiliar pairs). We defined several features based on the Spearman correlation and wavelet transform coherence (WTC) of biosignals measured on 23 pairs of subjects (13 familiar and 10 unfamiliar pairs) during eye contact for 10 min. Additional custom features that identify the maximum brain-to-body coupling instants between pairs were generated. After testing on combinations of different feature extraction methods, four subsets of features with the strongest discrimination power were taken into account to train a decision tree (DT) machine learning (ML) algorithm. We have obtained 95.65% classification accuracy using a leave-one-out cross-validation. The coupling features which represent the two maximum mean values resulting from the sum of 7 time-dependent WTC signals (oxyhaemoglobin concentration of the right prefrontal region, total haemoglobin concentration of the left and right prefrontal region, heart rate, electrodermal activity on the left and right wrist, and skin temperature on the right wrist) played an essential role in the classification accuracy. Training the DT-ML algorithm with combined brain and systemic physiology data provided higher accuracy than training it only with brain or systemic data alone. The results demonstrate the power of the SPA-fNIRS hyperscanning approach and the potential in applying ML to investigate the strength of social bonds in a wide range of social interaction contexts.

A Multiple Classification Model for the Prediction of Work-Related Musculoskeletal Disorder

In recent times, musculoskeletal disorders (MSD) represent one of the most common and expensive occupational health problems in both developed and developing countries. Work-related musculoskeletal disorders (WRMSD) are impairments that are mostly caused by the workplace and immediate environment. A two-step predictive model is introduced here using KNN and Decision tree machine learning algorithms. This model for predicting WRMSD enables for early detection and correction of upper and lower back disorders, carpal tunnel syndrome and other WRMSD disorders associated with office workers. Key informant interview technique, observation of previous methods, online repository and published related works were used in data gathering. In training the model, 80% of the dataset was used while 20% was used for testing the model to prevent overfitting using python programming language. JavaScript, Hypertext preprocessor (PHP), Hypertext Markup Language (HTML), Cascading Stylesheet (CSS) and MySQL were also used to develop the front and backend of the application. The result revealed that the proposed model had 90.44% accuracy, 92.71% Recall (sensitivity), 97.16% precision, and 94.88% F1-Score. The proposed model, however, makes it easy for multiple classifications in other to predict both present and future risk of WRMSD. Performance is estimated to have high accuracy, recall, precision and f1 score in comparison to other existing algorithms.

Machine Learning Model Generation With Copula-Based Synthetic Dataset for Local Differentially Private Numerical Data

With the development of IoT technology, personal data are being collected in many places. These data can be used to create new services, but consideration must be given to the individual’s privacy. We can safely collect personal data while adding noise by applying differential privacy. However, because such data are very noisy, the accuracy of machine learning trained by the data greatly decreased. In this study, our objective is to build a highly accurate machine learning model using these data. We focus on the decision tree machine learning algorithm, and, instead of applying it as is, we use a preprocessing technique wherein pseudodata are generated using a copula while removing the effect of noise added by differential privacy. In detail, the proposed novel protocol consists of three steps: generating a covariance matrix from the differentially private numerical data, generating a discrete cumulative distribution function from differentially private numerical data, and generating copula-based numerical samples. Simulation results using synthetic and real datasets verify the utility of the proposed method not only for the decision tree algorithm but also for other machine learning algorithms such as deep neural networks. This method will help create machine learning models, such as recommendation systems, using differential privacy data.

Predicting the dimensional variation of geometries produced through FDM 3D printing employing supervised machine learning

One of the most popular techniques of Additive Manufacturing (AM) being used currently is Fused Deposition Modelling (FDM). FDM is currently a nascent technology and has significant scope for improvement, particularly when it comes to the dimensional accuracy of the printed parts. The dimensional accuracy of the parts governs the usability of the product. These products when used in assemblies need to have very tight tolerances. These tolerances play a vital role in the quality of the finished product. For FDM to become commercially viable, the ability to predict this dimensional variation caused by different print parameters is essential, it would result in improved quality of the fabricated product and saving of time and resources. In this research, we studied the impact of significant parameters on the dimensional accuracy for different geometries such as cylindrical shafts, holes and rectangular slots. Predicting the dimensional variation is accomplished by using Decision Tree Machine Learning Algorithm. This algorithm's result is most suitable and provides accurate predictions; furthermore, the effectiveness of the model developed is validated by the R2 score of 0.67, the model can be further developed to establish industry-friendly functionality.

Identifying Coronary Artery Lesions by Feature Analysis of Radial Pulse Wave: A Case-Control Study.

Background Cardiovascular diseases have been always the most common cause of morbidity and mortality worldwide. Health monitoring of high-risk and suspected patients is essential. Currently, invasive coronary angiography is still the most direct and accurate method of determining the severity of coronary artery lesions, but it may not be the optimal clinical choice for suspected patients who had clinical symptoms of coronary heart disease (CHD) such as chest pain but no coronary artery lesion. Modern medical research indicates that radial pulse waves contain substantial pathophysiologic information about the cardiovascular and circulation systems; therefore, analysis of these waves could be a noninvasive technique for assessing cardiovascular disease. Objective The objective of this study was to analyze the radial pulse wave to construct models for assessing the extent of coronary artery lesions based on pulse features and investigate the latent value of noninvasive detection technology based on pulse wave in the evaluation of cardiovascular disease, so as to promote the development of wearable devices and mobile medicine. Method This study included 529 patients suspected of CHD who had undergone coronary angiography. Patients were sorted into a control group with no lesions, a 1 or 2 lesion group, and a multiple (3 or more) lesion group as determined by coronary angiography. The linear time-domain features and the nonlinear multiscale entropy features of their radial pulse wave signals were compared, and these features were used to construct models for identifying the range of coronary artery lesions using the k-nearest neighbor (KNN), decision tree (DT), and random forest (RF) machine learning algorithms. The average precision of these algorithms was then compared. Results (1) Compared with the control group, the group with 1 or 2 lesions had increases in their radial pulse wave time-domain features H2/H1, H3/H1, and W2 (P < 0.05), whereas the group with multiple lesions had decreases in MSE1, MSE2, MSE3, MSE4, and MSE5 (P < 0.05). (2) Compared with the 1 or 2 lesion group, the multiple lesion group had increases in T1/T (P < 0.05) and decreases in T and W1 (P < 0.05). (3) The RF model for identifying numbers of coronary artery lesions had a higher average precision than the models built with KNN or DT. Furthermore, average precision of the model was highest (80.98%) if both time-domain features and multiscale entropy features of radial pulse signals were used to construct the model. Conclusion Pulse wave signal can identify the range of coronary artery lesions with acceptable accuracy; this result is promising valuable for assessing the severity of coronary artery lesions. The technique could be used to development of mobile medical treatments or remote home monitoring systems for patients suspected or those at high risk of coronary atherosclerotic heart disease.

Detecting early signs of Alzheimer’s disease and related dementia onset from the EHR

As dementia is widely under-detected, a predictive model using electronic health records (EHR) could provide a method for early screening to implement preventive strategies. There is limited research on using EHR to identify persons with Alzheimer’s disease (AD) and related dementias (RD). In a data-driven approach, we used all ICD-9 diagnosis and CPT procedure codes from statewide inpatient, ambulatory surgery, and Medicare records, in addition to age at baseline and gender, to detect AD/RD from the Cache County Study on Memory in Aging (1995–2009). After removing participants diagnosed with dementia at baseline (n=335), 3882 (82%) Cache County Study participants could be linked to inpatient, ambulatory surgery, and/or Medicare EHR records; 484 (12.5%) of these 3882 had incident all-cause dementia, with 308 (7.9%) having AD/AD comorbid with RD; and 176 (4.5%) having RD without AD. We removed participant’s ICD-9 codes occurring after first AD/RD diagnoses. EHR features (~2000) along with gold-standard diagnoses as class labels were then used to train and detect AD and/or RD using a Gradient Boosting Trees machine learning algorithm. Models evaluated with nested cross-validation yielded AUCs of 0.70 for all-cause dementia, 0.69 for AD/AD comorbid with RD, and 0.67 for RD without AD. Key factors detecting AD/RD included age at enrollment, cardiovascular, metabolic, and kidney disease, and sleep disturbances, with feature importance varying by record type and time frame prior to dementia onset. Our findings suggest that a patient’s health status up to 12 years prior may be useful in identifying individuals at-risk for dementia development.

The Effects of Normalization and Standardization an Internet of Things Attack Detection

It is a known fact that we live in the computer age and that many devices in the world have access to the internet. So how secure are these devices? Is there any guarantee that user information is not accessed from intruder? After the concept of the Internet of Things came into our lives, many things such as seeing the food in our home refrigerator, connecting to the Internet from the car and, and video chatting from our smart watch entered our lives. The number of malicious software is also increasing with these new connections. Researchers are increasingly emphasizing the importance of network security and intensifying their studies. Data preprocessing is very important when designing a secure system. In this study, the importance of normalization and standardization in data preprocessing is examined to make machine learning approaches more successful for detecting attacks on IoT devices. The study was carried out in Logistic Regression, Decision Tree, and Stochastic Gradient Descent machine learning algorithms using the Bot-IoT dataset. Bot-IoT dataset is a popular dataset that is widely used in security studies on IoT devices. Normalization and standardization processes were applied to Bot-IoT dataset separately, so data preprocessing was performed, then selected machine learning algorithms were trained with these -normalized / standardized- datasets. As a result of the trainings made with machine learning algorithms, the values of Accuracy, Precision, Recall and F1 Score rates were examined. And as a result of the study, it was seen that the standardization increased the accuracy rate up to 99.96% in Logistic Regression.

A Proposal of Implementation of Sitting Posture Monitoring System for Wheelchair Utilizing Machine Learning Methods.

This paper presents a posture recognition system aimed at detecting sitting postures of a wheelchair user. The main goals of the proposed system are to identify and inform irregular and improper posture to prevent sitting-related health issues such as pressure ulcers, with the potential that it could also be used for individuals without mobility issues. In the proposed monitoring system, an array of 16 screen printed pressure sensor units was employed to obtain pressure data, which are sampled and processed in real-time using read-out electronics. The posture recognition was performed for four sitting positions: right-, left-, forward- and backward leaning based on k-nearest neighbors (k-NN), support vector machines (SVM), random forest (RF), decision tree (DT) and LightGBM machine learning algorithms. As a result, a posture classification accuracy of up to 99.03 percent can be achieved. Experimental studies illustrate that the system can provide real-time pressure distribution value in the form of a pressure map on a standard PC and also on a raspberry pi system equipped with a touchscreen monitor. The stored pressure distribution data can later be shared with healthcare professionals so that abnormalities in sitting patterns can be identified by employing a post-processing unit. The proposed system could be used for risk assessments related to pressure ulcers. It may be served as a benchmark by recording and identifying individuals’ sitting patterns and the possibility of being realized as a lightweight portable health monitoring device.

Interaction between the characteristics of urban three-dimensional landscape pattern and social-environmental factors based on boosted regression tree

Vertical expansion makes the structure and pattern of the city more complicated. Traditional two-dimensional landscape pattern cannot completely reflect the ecological structure and functional characteristics of urban landscape. In this study, we used the three-dimensional landscape pattern metrics to quantify the regional three-dimensional landscape pattern, and used boosted regression tree (BRT) machine learning algorithms to comprehensively analyze the interaction between social-environmental factors and urban landscape patterns in the central part of Shanghai. Results showed that high building ratio, mean architecture height, and architecture height standard deviation had higher values in the surrounding area of the inner ring. The number of buildings and landscape shape index were higher in the outer ring than those in other area. Building coverage ratio, floor area ratio and Shannon's diversity index had higher values in the central part, with the metrics of Puxi being generally higher than those of Pudong. Population density and normalized vegetation index (NDVI) interacted most significantly with the three-dimensional landscape pattern, with GDP as the least influential factor. Within a certain range, the three-dimensional landscape pattern metrics increased with larger population density in the social factors, and decreased with lower rate of NDVI and water surface ratio in the environmental factors. Our results demonstrated that the BRT method was effective in quantifying the interaction between landscape pattern and social-environmental factors. Our results help improve the understanding of the relationship between ecological environment and human well-being in the central part of Shanghai and provide a scientific basis for the urban three-dimensional expansion planning.

A machine learning approach for classifying bird and insect radar echoes with S-band Polarimetric Weather Radar

AbstractThe S-bandWSR-88D weather radar is sensitive enough to observe biological scatterers like birds and insects. However, their non-spherical shapes and frequent collocation in the radar resolution volume create challenges in identifying their echoes. We propose a method of extracting bird (or insect) features by coherently averaging dual polarization measurements from multiple radar scans, containing bird (insect) migration. Additional features are also computed to capture aspect and range dependence, and the variation of these echoes over local regions. Next, ridge classifier and decision tree machine learning algorithms are trained, first only with the averaged dual pol inputs and then different combinations of the remaining features are added. The performance of all models for both methods, are analyzed using metrics computed from the test data. Further studies on different patterns of birds/insects, including roosting birds, bird migration and insect migration cases, are used to further investigate the generality of our models. Overall, the ridge classifier using only dual polarization variables was found to perform consistently well across all these tests. Our recommendation is that this classifier can be used operationally on the US Next-Generation Radars (NEXRAD), as a first step in classifying biological echoes. It would be used in conjunction with the existing Hydrometeor Classification Algorithm (HCA), where the HCA would first separate biological from non-biological echoes, then our algorithm would be applied to further separate biological echoes into birds and insects. To the best of our knowledge, this study is the first to train a machine learning classifier that is capable of detecting diverse patterns of bird and insect echoes, based on dual polarization variables at each range gate.

Performance comparison of machine learning algorithms for data aggregation in social internet of things

Social Internet of Things (SIoT) is a paradigm of IoT where in objects are able to build social relationship among each other based on user preferences there by creating social platform. To establish relationship, diversified IoT devices interact and setup a connection between them. The relationship is built by considering same features, attribute, device type etc. The data generated by the heterogeneous devices of SIoT devices are huge and it has to be efficiently used. There exist few works related to data aggregation in SIoT in the literature. Hence, in this work a method is proposed to aggregate the SIoT data and conditions used to classify the relationship between the devices. The performance of Decision Tree (DT), K-Nearest Neighbors (KNN), Naive Bayes (NB) and Artificial Neural Network (ANN) machine learning algorithms are tested on the dataset. The experimental results show that DT and ANN algorithms performs well compared to other ML algorithms.

Predictive model construction for prediction of soil fertility using decision tree machine learning algorithm

Agriculture sector is recognized as the backbone of the Indian economy that plays a crucial role in the growth of the nation’s economy. It imparts on weather and other environmental aspects. Some of the factors on which agriculture is reliant are Soil, climate, flooding, fertilizers, temperature, precipitation, crops, insecticides, and herb. The soil fertility is dependent on these factors and hence difficult to predict. However, the Agriculture sector in India is facing the severe problem of increasing crop productivity. Farmers lack the essential knowledge of nutrient content of the soil, selection of crop best suited for the soil and they also lack efficient methods for predicting crop well in advance so that appropriate methods have been used to improve crop productivity. This paper presents different Supervised Machine Learning Algorithms such as Decision tree, K-Nearest Neighbor (KNN), Support Vector Machine (SVM) to predict the fertility of soil based on macro-nutrients and micro-nutrients status found in the dataset. Supervised Machine Learning algorithms are applied on the training dataset and are tested with the test dataset, and the implementation of these algorithms is done using R Tool. The performance analysis of these algorithms is done using different evaluation metrics like mean absolute error, cross-validation, and accuracy. Result analysis shows that the Decision tree is produced the best accuracy of 99% with a very less mean square error (MSE) rate.

Disease Prediction Using Machine Learning

Big data has a significant part in a number of businesses, but it is largely essential to the rapidly growing healthcare industry. It plays an important role by offering a large set of data points, constructing a robust system which allows for better and more accurate results in disease detection. Originally, the forecasts are made on the information accessible, but the absence of imperfect information contributes to a decrease in the caliber of precision. Besides incomplete data different qualities of particular regional diseases, which change based on their areas of origin can weaken the prediction models further. In this paper we use data mining techniques such as association rule mining, classification, clustering and finally the Decision Tree Machine learning algorithm to analyze the different kinds of general body-based illnesses. We implemented and assessed the efficacy of the Decision Tree algorithm over real-life clinical information.

Deciphering the Phillips Catalyst by Orbital Analysis and Supervised Machine Learning from Cr Pre-edge XANES of Molecular Libraries.

Unveiling the nature and the distribution of surface sites in heterogeneous catalysts, and for the Phillips catalyst (CrO3/SiO2) in particular, is still a grand challenge despite more than 60 years of research. Commonly used references in Cr K-edge XANES spectral analysis rely on bulk materials (Cr-foil, Cr2O3) or molecules (CrCl3) that significantly differ from actual surface sites. In this work, we built a library of Cr K-edge XANES spectra for a series of tailored molecular Cr complexes, varying in oxidation state, local coordination environment, and ligand strength. Quantitative analysis of the pre-edge region revealed the origin of the pre-edge shape and intensity distribution. In particular, the characteristic pre-edge splitting observed for Cr(III) and Cr(IV) molecular complexes is directly related to the electronic exchange interactions in the frontier orbitals (spin-up and -down transitions). The series of experimental references was extended by theoretical spectra for potential active site structures and used for training the Extra Trees machine learning algorithm. The most informative features of the spectra (descriptors) were selected for the prediction of Cr oxidation states, mean interatomic distances in the first coordination sphere, and type of ligands. This set of descriptors was applied to uncover the site distribution in the Phillips catalyst at three different stages of the process. The freshly calcined catalyst consists of mainly Cr(VI) sites. The CO-exposed catalyst contains mainly Cr(II) silicates with a minor fraction of Cr(III) sites. The Phillips catalyst exposed to ethylene contains mainly highly coordinated Cr(III) silicates along with unreduced Cr(VI) sites.

227 Poor sleep as a predictor of COVID-19 related stress, fear and sadness in young adolescents: a longitudinal study

IntroductionAdolescence is a transitional life-stage accompanied by large biopsychosocial changes and greater psychophysiological vulnerability. Global events like the COVID-19 pandemic may increase vulnerability to depression and anxiety in this population. Poor sleep is often associated with depression, and both sleep and mood have been shown to be strongly impacted by the COVID-19 pandemic, with most studies focusing on adults. The current study investigates psychological distress in young adolescents during the pandemic, and specifically, whether poor sleep before the pandemic predicts psychological distress.MethodsSelf-report data were analyzed from 3099 adolescents (9-10 years at baseline) in the population-based, demographically diverse, Adolescent Brain Cognitive Development (ABCD) study across three pre-pandemic annual visits and 3 monthly time points during the COVID-19 pandemic (ages 11-13 years). At each assessment, children and their guardians completed questionnaires including those about sleep, environment, and psychological wellbeing. Gradient Boosted Tree machine learning algorithms were used to identify the strongest predictors of pandemic-related psychological distress in individuals. We trained models using pre-pandemic sleep measures along with demographics, economic, and social measures during the pandemic. We evaluated the performance of the models using area under curve (AUC) metrics and interpreted the models by using the recently proposed SHapley Additive exPlanations methodology.ResultsPandemic-related perceived stress, fear and sadness were accurately detected with our classifiers (AUC = 0.83 for perceived stress, AUC = 0.73 for fear, AUC = 0.79 for sadness). Across all models, shorter sleep duration, prolonged sleep onset latency, and longer time between waking and getting out of bed predicted greater distress. Moreover, female sex, and pandemic-related factors, including greater family conflict, fewer economic resources, and more screen time contributed to prediction performance in all three models.ConclusionFindings highlight the importance of addressing sleep problems and ensuring sufficient sleep duration in children to protect against the psychological impact of major life events, including the COVID-19 pandemic. Considering the long-lasting effects of sleep, it would be crucial to improve sleep health by targeted prevention, intervention and increased awareness among adolescents.Support (if any)NIH U01DA041022

Incorporating real-world evidence into the development of patient blood glucose prediction algorithms for the ICU

ObjectiveGlycemic control is an important component of critical care. We present a data-driven method for predicting intensive care unit (ICU) patient response to glycemic control protocols while accounting for patient heterogeneity and variations in care.Materials and MethodsUsing electronic medical records (EMRs) of 18 961 ICU admissions from the MIMIC-III dataset, including 318 574 blood glucose measurements, we train and validate a gradient boosted tree machine learning (ML) algorithm to forecast patient blood glucose and a 95% prediction interval at 2-hour intervals. The model uses as inputs irregular multivariate time series data relating to recent in-patient medical history and glycemic control, including previous blood glucose, nutrition, and insulin dosing.ResultsOur forecasting model using routinely collected EMRs achieves performance comparable to previous models developed in planned research studies using continuous blood glucose monitoring. Model error, expressed as mean absolute percentage error is 16.5%–16.8%, with Clarke error grid analysis demonstrating that 97% of predictions would be clinically acceptable. The 95% prediction intervals achieve near intended coverage at 93%–94%.DiscussionML algorithms built on observational data sources, such as EMRs, present a promising approach for personalization and automation of glycemic control in critical care. Future research may benefit from applying a combination of methodologies and data sources to develop robust methodologies that account for the variations seen in ICU patients and difficultly in detecting the extremes of observed blood glucose values.ConclusionWe demonstrate that EMRs can be used to train ML algorithms that may be suitable for incorporation into ICU decision support systems.

Development of a computerized adaptive diagnostic screening tool for psychosis

We develop a two-stage diagnostic classification system for psychotic disorders using an extremely randomized trees machine learning algorithm. Item bank was developed from clinician-rated items drawn from an inpatient and outpatient sample. In stage 1, we differentiate schizophrenia and schizoaffective disorder from depression and bipolar disorder (with psychosis). In stage 2 we differentiate schizophrenia from schizoaffective disorder. Out of sample classification accuracy, determined by area under the receiver operator characteristic (ROC) curve, was outstanding for stage 1 (Area under the ROC curve (AUC) = 0.93, 95% confidence interval (CI) = 0.89, 0.94), and excellent for stage 2 (AUC = 0.86, 95% CI = 0.83, 0.88). This is achieved based on an average of 5 items for stage 1 and an average of 6 items for stage 2, out of a bank of 73 previously validated items.

Spam email detection using machine learning algorithm

Nowadays the communication between the organizations or any individuals became easier by use of Electronic mail method. The internet users are increasing rapidly day by day and also spams are increasing with the emails. It was an easy task for the spammers to create an email account and making a fake profile. Therefore, detecting of these spam mails that were fraud is of most important. This paper aims to develop a proposed approach of data science for spam email detection (SMD) using machine learning algorithm. A hybrid bagging approach is used in this proposed method for the detection of spam emails which implements the two, Naïve Bayes and J48 (i.e. decision tree) machine learning algorithms. Each of these algorithms is applied as an input with a data set which is partitioned in to different sets using the data science. Emails classification can be accomplished depends on the patterns of repetitive keywords and several additional parameters like Cc (carbon copy) or Bcc (Blind carbon copy), domain, header etc. that are present in their structure. Whenever the machine learning algorithm is applied for the email classification every one of those parameters are considered as a features. This proposed approach of data science for spam mail detection using machine learning algorithm achieved a 88.12% of overall accuracy with the hybrid bagged approach implementation.

Prediction of lameness using automatically recorded activity, behavior and production data in post-parturient Irish dairy cows

BackgroundAlthough visual locomotion scoring is inexpensive and simplistic, it is also time consuming and subjective. Automated lameness detection methods have been developed to replace the visual locomotion scoring and aid in early and accurate detection. Several types of sensors are measuring traits such as activity, lying behavior or temperature. Previous studies on automatic lameness detection have been unable to achieve high accuracy in combination with practical implementation in a on farm commercial setting. The objective of our research was to develop a prediction model for lameness in dairy cattle using a combination of remote sensor technology and other animal records that will translate sensor data into easy to interpret classified locomotion information for the farmer. During an 11-month period, data from 164 Holstein-Friesian dairy cows were gathered, housed at an Irish research farm. A neck-mounted accelerometer was used to gather behavioral metrics, additional automatically recorded data consisted of milk production and live weight. Locomotion scoring data were manually recorded, using a one-to-five scale (1 = non-lame, 5 = severely lame). Locomotion scores where then used to label the cows as sound (locomotion score 1) or unsound (locomotion score ≥ 2). Four supervised classification models, using a gradient boosted decision tree machine learning algorithm, were constructed to investigate whether cows could be classified as sound or unsound. Data available for model building included behavioral metrics, milk production and animal characteristics.ResultsThe resulting models were constructed using various combinations of the data sources. The accuracy of the models was then compared using confusion matrices, receiver-operator characteristic curves and calibration plots. The model which achieved the highest performance according to the accuracy measures, was the model combining all the available data, resulting in an area under the curve of 85% and a sensitivity and specificity of 78%.ConclusionThese results show that 85% of this model’s predictions were correct in identifying cows as sound or unsound, showing that the use of a neck-mounted accelerometer, in combination with production and other animal data, has potential to replace visual locomotion scoring as lameness detection method in dairy cows.

Applications of Artificial Intelligence for Analysis of Two-Phase Flow in PEM Fuel Cell Flow Fields

Many electrochemical devices rely upon flow-field channels to transport reactants and byproducts to and from critical electrode sites. Often, this results in multiphase flow, which, if not managed properly, can reduce system performance and lifespan. In the case of PEM fuel cells, two-phase flow occurs when liquid water forms as a byproduct of the cathode oxygen reduction reaction. Over the past two-decades much effort has been invested to understand the nature of this stochastic transport phenomena and how to optimize flow-field design to mitigate its effects. This work probes the use of artificial intelligence to guide our understanding of complex relationships between liquid phase behavior and pressure drop in these reactant channels [1]. Using a supervised decision tree (DT) machine learning algorithm, the liquid distributions in ex-situ experimental reactant channels are correlated to the induced pressure drop for a range of air flow rates; this process is diagrammed in Fig. 1. Liquid phase data collection is accomplished by capturing 2D images via a CCD camera of liquid water distributions in a transparent flow channel simulating PEM fuel cell conditions. In this setup, liquid water is injected through the bottom of the GDL in the same manner that water must travel from the catalyst layer through the GDL to reach the channel an operating cell. Pressure transducers measure pressure drop along the length of the channel and are synchronized with the image collection. The images are processed for noise reduction and feature extraction and then used with pressure data to train a regression DT. The models produced using the DT show the potential to improve predictions over earlier methods that relied solely upon liquid area [2]. Overall, the DT models predicted pressure drop using liquid distributions with 88.9% to 98.54% accuracy and were able to capture complex behavior such as pressure changes due to the breakup of liquid slugs. In general, the methods may be valuable for attaining deeper understanding of the impact variables such as geometry, hydrophobicity, and system operating conditions have on two-phase flow pressure drop trends. Additionally, these models may serve as a validation tool for numerical efforts aimed at flow-field design. A reliable model that can accurately predict the liquid-gas two-phase flow pressure drop may be used to estimate the amount and potentially the distribution of the water content within the flow channel by inputting the pressure drop measured during the operation of the cell.Figure 1: Overview of the regression DT training and testing processes.[1] Santamaria AD, Mortazavi M, Chauhan V, Benner J. Two-Phase Flow Characterization in PEM Fuel Cells Using Machine Learning. In Meeting s 2019 May 1 (No. 30, pp. 1538-1538). The Electrochemical Society.[2] Hussaini IS, Wang CY. Visualization and quantification of cathode channel flooding in PEM fuel cells. Journal of Power Sources. 2009 Feb 15;187(2):444-51. Figure 1