No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Predicting Heatstroke Risk and Preventing Health Complications Using Machine Learning and Physiological Data
Abstract
The rise in global temperatures has become a significant concern, leading to an increase in heat stroke incidents, which pose severe
health consequences, including mortality. The comfort levels of indoor environments fluctuate depending on various activities
performed in different situations. Physiological data, encompassing heart rate, body temperature, and blood pressure, provide valuable
insights into the identification of patterns and trends that may signify an elevated risk of heatstroke. However, manual analysis of such
data proves impractical due to its complexity and volume. In this paper, we present an energy-efficient machine learning-based
approach to forecast individual thermal comfort sensations, enabling the early identification of individuals at risk of heatstroke before
symptom manifestation. We conducted experiments using four distinct machine learning models along with one deep learning-based
model, achieving an accuracy of approximately 99% on test set. Code is available at
https://github.com/mdhosen/Heatstroke-prevention.
ResearchGate has not been able to resolve any citations for this publication.
Background:
Accurate estimation of historical PM2.5 (particle matter with an aerodynamic diameter of less than 2.5μm) is critical and essential for environmental health risk assessment.
Objectives:
The aim of this study was to develop a multiple-level stacked ensemble machine learning framework for improving the estimation of the daily ground-level PM2.5 concentrations.
Methods:
An innovative deep ensemble machine learning framework (DEML) was developed to estimate the daily PM2.5 concentrations. The framework has a three-stage structure: At the first stage, four base models [gradient boosting machine (GBM), support vector machine (SVM), random forest (RF), and eXtreme gradient boosting (XGBoost)] were used to generate a new data set of PM2.5 concentrations for training the next-stage learners. At the second stage, three meta-models [RF, XGBoost, and Generalized Linear Model (GLM)] were used to estimate PM2.5 concentrations using a combination of the original data set and the predictions from the first-stage models. At the third stage, a nonnegative least squares (NNLS) algorithm was employed to obtain the optimal weights for PM2.5 estimation. We took the data from 133 monitoring stations in Italy as an example to implement the DEML to predict daily PM2.5 at each 1km×1km grid cell from 2015 to 2019 across Italy. We evaluated the model performance by performing 10-fold cross-validation (CV) and compared it with five benchmark algorithms [GBM, SVM, RF, XGBoost, and Super Learner (SL)].
Results:
The results revealed that the PM2.5 prediction performance of DEML [coefficients of determination (R2)=0.87 and root mean square error (RMSE)=5.38μg/m3] was superior to any benchmark models (with R2 of 0.51, 0.76, 0.83, 0.70, and 0.83 for GBM, SVM, RF, XGBoost, and SL approach, respectively). DEML displayed reliable performance in capturing the spatiotemporal variations of PM2.5 in Italy.
Discussion:
The proposed DEML framework achieved an outstanding performance in PM2.5 estimation, which could be used as a tool for more accurate environmental exposure assessment. https://doi.org/10.1289/EHP9752.
Thermal comfort is defined as the condition of mind that expresses satisfaction with the thermal environment and it is assessed by subjective evaluation. Achieving comfortable environments is essential for human health, productivity, learning performance and energy efficiency. The measurement of thermal comfort requires an indirect process involving several domains: thermal environment, physiology and psychology. Four physiological signals were reviewed, considering their relevance in the context of measuring indoor thermal comfort. The measurement approaches for investigating electroencephalography, electrocardiograph, skin temperature and galvanic skin response signals in the field of thermal comfort are presented. This paper introduces an overview on the application of wearable sensors for recording physiological parameters and extracting features potentially correlated with thermal comfort, together with a discussion about their reliability. The review shows the state of the art, the identification of existing knowledge gaps in this area and the corresponding needs for future research and dedicated methodological efforts.
Various approaches have been proposed to model PM 2.5 in the recent decade, with satellite-derived aerosol optical depth, land-use variables, chemical transport model predictions, and several meteorological variables as major predictor variables. Our study used an ensemble model that integrated multiple machine learning algorithms and predictor variables to estimate daily PM 2.5 at a resolution of 1 km × 1 km across the contiguous United States. We used a generalized additive model that accounted for geographic difference to combine PM 2.5 estimates from neural network, random forest, and gradient boosting. The three machine learning algorithms were based on multiple predictor variables, including satellite data, meteorological variables, land-use variables, elevation, chemical transport model predictions, several reanalysis datasets, and others. The model training results from 2000 to 2015 indicated good model performance with a 10-fold cross-validated R 2 of 0.86 for daily PM 2.5 predictions. For annual PM 2.5 estimates, the cross-validated R 2 was 0.89. Our model demonstrated good performance up to 60 μg/m 3. Using trained PM 2.5 model and predictor variables, we predicted daily PM 2.5 from 2000 to 2015 at every 1 km × 1 km grid cell in the contiguous United States. We also used localized land-use variables within 1 km × 1 km grids to downscale PM 2.5 predictions to 100 m × 100 m grid cells. To characterize uncertainty, we used meteorological variables, land-use variables, and elevation to model the monthly standard deviation of the difference between daily monitored and predicted PM 2.5 for every 1 km × 1 km grid cell. This PM 2.5 prediction dataset, including the downscaled and uncertainty predictions, allows epidemiologists to accurately estimate the adverse health effect of PM 2.5. Compared with model performance of individual base learners, an ensemble model would achieve a better overall estimation. It is worth exploring other ensemble model formats to synthesize estimations from different models or from different groups to improve overall performance.
Thermal comfort is an essential environmental factor related to quality of life and work effectiveness. We assessed the feasibility of wrist skin temperature monitoring for estimating subjective thermal sensation. We invented a wrist band that simultaneously monitors skin temperatures from the wrist (i.e., the radial artery and ulnar artery regions, and upper wrist) and the fingertip. Skin temperatures from eight healthy subjects were acquired while thermal sensation varied. To develop a thermal sensation estimation model, the mean skin temperature, temperature gradient, time differential of the temperatures, and average power of frequency band were calculated. A thermal sensation estimation model using temperatures of the fingertip and wrist showed the highest accuracy (mean root mean square error [RMSE]: 1.26 ± 0.31). An estimation model based on the three wrist skin temperatures showed a slightly better result to the model that used a single fingertip skin temperature (mean RMSE: 1.39 ± 0.18). When a personalized thermal sensation estimation model based on three wrist skin temperatures was used, the mean RMSE was 1.06 ± 0.29, and the correlation coefficient was 0.89. Thermal sensation estimation technology based on wrist skin temperatures, and combined with wearable devices may facilitate intelligent control of one's thermal environment.
As a result of climate change, the frequency of extreme temperature events is expected to increase and such events are associated with increased morbidity and mortality. Vulnerability patterns, and corresponding adaptation strategies, are most usefully conceptualized at a local level.
We used a case-only analysis to examine subject and neighborhood characteristics that modified the association between heat waves and mortality. All deaths of New York City residents from 2000 through 2011 were included in this analysis. Meteorological data was obtained from the National Climatic Data Center. Modifying characteristics were obtained from the death record and geographic data sets.
A total of 234,042 adult deaths occurred during the warm season of our study period. Compared to other warm season days, deaths during heat waves were more likely to occur in black (non-Hispanic) individuals than other race/ethnicities (odds ratio (OR) = 1.08, 95% confidence interval (CI), 1.03, 1.12), more likely to occur at home than in institutions and hospital settings, OR = 1.11 (95% CI: 1.06, 1.16), and more likely among those living in census tracts that received greater public assistance, OR = 1.05 (95% CI: 1.01, 1.09). Finally, deaths during heat waves were more likely among residents in areas of the city with higher relative daytime summer surface temperature and less likely among residents living in areas with more green space.
Mortality during heat waves varies widely within a city. Understanding which individuals and neighborhoods are most vulnerable can help guide local preparedness efforts.
Climate models project that heat waves will increase in frequency and severity. Despite many studies of mortality from heat waves, few studies have examined morbidity.
In this study we investigated whether any age or race/ethnicity groups experienced increased hospitalizations and emergency department (ED) visits overall or for selected illnesses during the 2006 California heat wave.
We aggregated county-level hospitalizations and ED visits for all causes and for 10 cause groups into six geographic regions of California. We calculated excess morbidity and rate ratios (RRs) during the heat wave (15 July to 1 August 2006) and compared these data with those of a reference period (8-14 July and 12-22 August 2006).
During the heat wave, 16,166 excess ED visits and 1,182 excess hospitalizations occurred statewide. ED visits for heat-related causes increased across the state [RR = 6.30; 95% confidence interval (CI), 5.67-7.01], especially in the Central Coast region, which includes San Francisco. Children (0-4 years of age) and the elderly (> or = 65 years of age) were at greatest risk. ED visits also showed significant increases for acute renal failure, cardiovascular diseases, diabetes, electrolyte imbalance, and nephritis. We observed significantly elevated RRs for hospitalizations for heat-related illnesses (RR = 10.15; 95% CI, 7.79-13.43), acute renal failure, electrolyte imbalance, and nephritis.
The 2006 California heat wave had a substantial effect on morbidity, including regions with relatively modest temperatures. This suggests that population acclimatization and adaptive capacity influenced risk. By better understanding these impacts and population vulnerabilities, local communities can improve heat wave preparedness to cope with a globally warming future.
Ceiling fans are common in Indian residences as airflow is important for thermal comfort. Ceiling fans are generally located in the center of room for good distribution of airflow though airflow is limited to a small area and air velocity is reduced with increasing distance from the fan. Occupants spatially adapt while deciding seating layout considering various functional factors including thermal comfort. This study evaluates the effect of ceiling fan induced non uniform airflow on thermal comfort conditions in four living room seating layouts. The thermal environment was monitored in a living room and ceiling fan induced airflow was measured to study the distribution of airflow in four selected furniture layouts. The thermal environment in seats varied widely due to non-uniform distribution of airflow in the space. Thermal sensations in 12 seating positions of 4 different layouts were analyzed using Tropical Summer Index (TSI) and Predicted Mean Vote (PMV). Thermal comfort conditions in different seats improved at higher fan speeds in all layouts. TSI and PMV based thermal sensations varied from ‘Too hot’ to ‘Slightly Cool’ between warm summer and cool winter tropical seasons. The number of seats with ‘Too Hot’ thermal sensation were reduced at higher fan speed. Overall, seats farther from the fan experience relatively hotter thermal sensation in summer months and are not affected by fan speeds. The study emphasizes the relevance of moveable furniture and role of spatial adaptive behavior in room layout design, where required airflow for thermal comfort is not available within the space.
Hybrid energy systems that contain renewable energy such as wind and solar energy
help to minimize CO2 emissions. Therefore, studying these systems to improve their
performance has become one of the critical needs these days due to the
environmental crisis. Within HES, the scheduling and managing of energy sources is
an essential topic that has been covered in detail by numerous studies, as errors in
SAM can lead to HES blackouts. Recent research has experimented with SAM
software to achieve optimal SAM. This work presents two objectives. The first objective
is to forecast which energy source should supply the demand side by applying different
machine-learning algorithms such as Random Forest, Decision Tree, Gaussian Naive
Bayes and K-Nearest Neighbors. The second objective is to compare the results of
these algorithms to choose the best algorithm for the best performance, and rank them
based on their performance and try to improve their accuracy. The work is validated by
using different algorithms. The results show that DT algorithm has achieved the best
performance compared to the RF and Gaussian NB algorithms. KNN algorithm gives
the lowest accuracy especially over class 3. The results proof that RF, DT, and
Gaussian NB algorithms are reliable.
Aiming at the problem of the safety monitoring of the surrounding environment of the glass window on the building, a scheme of distributed optical fiber acoustic sensors based on the phase sensitive time-domain reflection is proposed. Firstly, Wigner bispectrum is used to analyze the time–frequency energy distribution of vibration signal, then the axially integrated bispectrum algorithm is used to extract the time–frequency characteristic components, and the method of sub-band division is used to avoid information loss and decrease the cost of computing. The eigenvectors of different vibration events are recognized by extreme gradient boosting tree algorithm. Based on the diversity of acoustic signal types in the environment, there are eight kinds of events including wind blowing, knocking, window opening, watering, hammering, dog barking, aircraft sound and no disturbance to be selected for vibration experiments. The experimental results show that the scheme can get better recognition effect, and the average recognition rate of eight kinds of events reaches 93.3%.
In this chapter, we explore Support Vector Machine (SVM)-a machine learning method that has become exceedingly popular for neuroimaging analysis in recent years. Because of their relative simplicity and flexibility for addressing a range of classification problems, SVMs distinctively afford balanced predictive performance, even in studies where sample sizes may be limited. In brain disorders research, SVMs are typically employed using multivoxel pattern analysis (MVPA) because their relative simplicity carries a lower risk of overfitting even using high-dimensional imaging data. More recently, SVMs have been used in the context of precision psychiatry, particularly for applications that involve predicting diagnosis and prognosis of brain diseases such as Alzheimer’s disease, schizophrenia, and depression. In the last section of this chapter, we review a number of recent studies that use SVM for such applications.
Extreme heat events have recently become more frequent and represent an increasing risk of damage to public health. However, the existing prediction of heatwave related health effects has limited representativeness and verification. Our study addressed the prediction of heatstroke occurrences based on three years' data of typical cities of hot temperature in China, and examined the importance ranks of model parameters including meteorological and socioeconomic status (SES) factors. The results show that meteorological factors contributed the most to model estimation of the parameters evaluated, and SES parameters, such as the search index, were also important indicators of heatstroke prediction. The model had a satisfying performance compared to traditional linear regression models. The model established in our study can be further applied to extreme weather-related impact research and reduce economic loss due to public health expenses.
While previous studies have attempted to understand and predict users' behaviors and preferences for residential thermostats, they have been restricted by a lack of available data. Because of practical constraints, researchers previously relied on short observation periods, small sample groups, and/or participants close in physical location. The advent of the connected thermostat and its inherent centralized data collection now allows for such studies to be performed without the onus of data collection. Specifically, in this article we focus on the ‘Donate Your Data’ dataset made available by the thermostat manufacturer ecobee Inc. The dataset, consisting of more than 10,000 connected thermostats installed across North America and spanning multiple years, was used to investigate how users' comfort decisions are affected by exterior stimuli such as climate regions, seasonal patterns, and utility rates. Our analysis indicates that seasonality and climate region affected user preferences while utility rates did not contribute to meaningful variation in behavior. Further investigation explored if behavioral user types could be identified based on variation in occupied and unoccupied setpoints, thermostat overrides with holds, or heating and cooling setpoint selection. We did not find distinct user clusters to be identifiable based on any of the metrics; rather, occupant behavior in the population appeared to span more of a continuum across each metric.
In practice, building heating, ventilation, and air conditioning (HVAC) systems are essentially set at nominal levels according to industry guidelines. However, several studies have demonstrated that this conventional practice is unlikely to meet the thermal requirements of occupants in a single or multi-occupancy space due to occupants' diverse preferences, activities and needs. To improve occupants' thermal comfort, this study develops and tests a smartphone application framework which is capable of dynamically determining the optimum room conditioning mode (mechanical conditioning or natural ventilation) and HVAC settings (thermostat setpoint) in single and multi-occupancy spaces. The “personalized” HVAC control framework integrates environment data (obtained from sensors) with human physiological and behavioral data (obtained from wearable devices, polling apps) in a smartphone application we developed for human-building interaction. In the operation phase, occupants' thermal preferences are continuously predicted using the personalized comfort models, developed from the training data through the Random Forest classifier, when determining the optimum HVAC control strategies. Two case studies are conducted to demonstrate the capabilities of the developed framework to improve thermal comfort in single and multi-occupancy spaces.
This paper reviews the most used thermal comfort models and indicators with their variants, discussing their usage in control problems referring to energy management in indoor applications. The first part addresses the recent literature referring to the thermal comfort concepts, models of human thermal comfort, thermal comfort models and indicators, thermal comfort standards, control systems, optimisation methods, and practical assessments. Then, the ambient and personal parameters used to represent thermal comfort and thermal sensation are recalled. The following part reviews the definitions and usage of a number of thermal comfort indices, mainly related to the Predicted Mean Vote (PMV), the Actual Mean Vote (AMV), and the Predicted Percentage Dissatisfied (PPD), with their modifications and variants, indicating a number of applications to different situations in indoor environments. The last part reviews the thermal comfort models used to define control strategies in indoor applications, discussing the characteristics and parameters of models based on artificial neural networks, autoregressive variants, fuzzy control, and hybrid models combining different approaches. The characteristics of these models and their usage to predict the indoor air temperature and the PMV index are discussed with reference to the identification of the several inputs used in relevant literature contributions.
Inefficient controlling strategies in heating and cooling systems have given rise to a large amount of energy waste and to widespread complaints about the thermal environment in buildings. An intelligent control method based on a support vector machine (SVM) classifier is proposed in this paper. Skin temperatures are the only inputs to the model and have shown attractive prediction power in recognizing steady state thermal demands. Data were accumulated from two studies to consider potential use for either individuals or a group of occupants. Using a single skin temperature correctly predicts 80% of thermal demands. Using combined skin temperatures from different body segments can improve the model to over 90% accuracy. Results show that three skin locations contained enough information for classification and more would cause the curse of dimensionality. Models using different skin temperatures were compared. Optimal parameters for each model were provided using grid search technique. Considering the overfitting possibility and the cases without learning processes, SVM classifiers with a linear kernel are preferred over Gaussian kernel ones.
High on the political and social agenda are the levels of domestic energy use, of carbon emissions, and of fuel poverty. Strategies to reduce these levels should ensure that the health and well-being of residents are protected and improved. Thermal discomfort in housing ranges from being uncomfortable to serious health impacts. Avoiding thermal discomfort maintains well-being and protects the health of, in particular, the most susceptible members of the population. This position paper defines thermal discomfort, examines how it can be measured, and looks at the health impacts of low and high temperatures. Factors that increase susceptibility to low and to high temperatures, both human and building related, are identified and discussed. This paper shows that there are similarities between some of these factors, and that tackling these together will provide protection for those most at risk from exposure to high and low temperatures. Strategies aimed at reducing susceptibility to both excess heat and excess cold are suggested.
This work shows some of the results of a field study about environmental comfort investigations in classrooms. In this project thermal, acoustic, visual and air quality aspects were analysed in a number of classrooms—13 classrooms at four different high schools of the Provincia di Torino and four typical medium-sized university classrooms of the Politecnico di Torino, Italy. The investigations were carried out during the heating period. Both field measurements and subjective surveys were performed at the same time during the regular lesson periods.This paper focuses on thermal comfort, which may have a significant effect on the students’ performance, in terms of attention, comprehension and learning levels.The measurement campaign consisted in measuring the thermal environment parameters—air temperature, mean radiant temperatures, air relative humidity and air velocity. Through these data, the thermal comfort Fanger's indices (predicted mean vote (PMV), and predicted percentage of dissatisfied (PPD)people) were calculated, the actual people clothing and metabolic rate being known.The subjective survey involved questions on the thermal environmental perception. They basically investigated the thermal environment acceptability and preference. Moreover, a judgement based on the typical seven point thermal sensation scale (Fanger 7-points scale) was also asked. Through the elaboration of the questionnaire data, the actual percentage of dissatisfied (PD) people of the felt thermal environment was evaluated.The judgements about the thermal environment were compared with the results of the field measurements. Moreover, the subjective mean votes were compared with the thermal environment perceptions in terms of acceptability and preference.
It requires more energy to maintain a narrow indoor temperature range than a broader range, in which the building may be allowed to float with reduced conditioning for longer periods of time. A narrow range should presumably be preferable to the building occupants to justify its increased energy cost. At what widths are temperature ranges detected, preferred, or judged unacceptable? Three databases of occupant satisfaction in buildings are used to examine the acceptability of three classes of temperature range currently employed in the ISO and European standards, and proposed for the ASHRAE standard. These are alternatively identified as class A, B, and C, or category I, II, and III, but their specifications are identical. The A class (I category) is found to confer no relative satisfaction benefit to individuals or to realistic building occupancies. In addition, the differences in B and C class satisfaction are small.
Web 2.0 technologies, such as wikis, blogs, tags and feeds, have been adopted and adapted by software engineers. With the annual Web2SE workshop, we provide a venue for research on Web 2.0 for software engineering by highlighting state-of-the-art work, ...
Accelerometers are promising tools for characterizing physical activity (PA) patterns in free-living persons. To date, validation of energy expenditure (EE) predictions from accelerometers has been restricted to short laboratory or simulated free-living protocols. This study seeks to determine the capabilities of eight previously published regression equations for three commercially available accelerometers to predict summary measures of daily EE.
Study participants were outfitted with ActiGraph, Actical, and RT3 accelerometers, while measurements were simultaneously made during overnight stays in a room calorimeter, which provided minute-by-minute EE measurements, in a diverse subject population (n = 85). Regression equations for each device were used to predict the minute-by-minute metabolic equivalents (METs) along with the daily PA level (PAL).
Two RT3 regressions and one ActiGraph regression were not significantly different from calorimeter measured PAL. When data from the entire visit were divided into four intensity categories-sedentary, light, moderate, and vigorous-significant (P < 0.001) over- and underpredictions were detected in numerous regression equations and intensity categories.
Most EE prediction equations showed differences of <2% in the moderate and vigorous intensity categories. These differences, though small in magnitude, may limit the ability of these regressions to accurately characterize whether specific PA goals have been met in the field setting. New regression equations should be developed if more accurate prediction of the daily PAL or higher precision in determining the time spent in specific PA intensity categories is desired.
- X Huang
- A Khetan
- M Cvitkovic
- Z Karnin
Huang, X., Khetan, A., Cvitkovic, M., Karnin, Z.: Tabtransformer: Tabular data modeling using contextual embeddings. arXiv preprint arXiv:2012.06678 (2020)
Validation of the empatica e4 wristband
- C Mccarthy
- N Pradhan
- C Redpath
- A Adler
McCarthy, C., Pradhan, N., Redpath, C., Adler, A.: Validation of the empatica e4
wristband. In: 2016 IEEE EMBS international student conference (ISC), pp. 1-4.
IEEE (2016)
Drought and heatwaves in europe: Historical reconstruction and future projections
- L Samaniego
- S Thober
- R Kumar
- O Rakovec
- E Wood
- J Sheffield
- M Pan
- N Wanders
- C Prudhomme
Samaniego, L., Thober, S., Kumar, R., Rakovec, O., Wood, E., Sheffield, J., Pan, M.,
Wanders, N., Prudhomme, C.: Drought and heatwaves in europe: Historical reconstruction and future projections. In: EGU General Assembly Conference Abstracts, p. 5659
(2017)
Toward the prediction of environmental thermal comfort sensation using wearables
- K H T Hossain
- Y Kawasaki
- G Lopez
T Hossain, K.H., Kawasaki, Y., Lopez, G.: Toward the prediction of environmental
thermal comfort sensation using wearables. Workshops at 18th International Conference on Intelligent Environments (IE2022) (2022)
5th abc challenge: Forecasting thermal comfort sensations for heatstroke prevention: Leveraging physiological data for better outcomes
- K N T Hossain
- K H Messaoudi
- C G T Tazin
- G Lopez
T Hossain, K.N., I E Messaoudi, K.H., Tazin, C.G.T., Lopez, G.: 5th abc challenge: Forecasting thermal comfort sensations for heatstroke prevention: Leveraging physiological data for better outcomes (2023). DOI 10.21227/ys7g-6t64. URL
https://dx.doi.org/10.21227/ys7g-6t64