Random forests are a cornerstone of ensemble learning due to their robustness and scalability. However, their potential is constrained by a rigid voting structure. By assigning equal weight to every tree, the standard model overlooks the heterogeneous performance inherent to individual trees across diverse target classes. In this paper, we introduce the Fractional Class-Specific Weighted Random Forest (fCWRFO). This approach refines tree level confidence weights through fractional order gradient descent. Indeed, it expands the optimization landscape beyond standard integer derivatives which only represent a specific instance of our framework (α=1). This mechanism introduces a history-dependent fractional scaling of the weight updates, which adaptively modulates the optimization trajectory and improves convergence stability. Comprehensive benchmarks on 40 UCI datasets demonstrate that fCWRFO improves generalization accuracy and stability, outperforming both standard Random Forests and classical integer-based weighting schemes.

Constructing optimal decision trees is an NP-complete problem. To address this challenge, we propose PSO-LWFWDT, a framework that reformulates tree induction as a continuous optimization problem. The proposed approach introduces a continuous level-wise feature-weighting mechanism optimized through Particle Swarm Optimization (PSO). Gini gain is dynamically modulated by these weights, while a hard-thresholding mechanism suppresses weak attributes, thereby enabling implicit depth-specific feature selection. The learning process is guided by a multi-objective fitness function that maximizes the weighted F1-score while penalizing tree depth, encouraging the induction of parsimonious trees without the need for post-pruning. An extensive evaluation on 30 benchmark datasets shows that PSO-LWFWDT achieves competitive or superior predictive performance compared with CART, Oblique Tree, and NCART, while maintaining controlled model complexity.

Background Patients with COVID‐19 are at risk of thrombotic and cardiovascular complications. Existing risk estimates derive from early COVID‐19 experiences. More recent event rates, temporal trends, and the association between vaccination status and cardiovascular outcomes remain unclear. Methods The CORONA‐VTE Network (COVID‐19 Registry to Assess Frequency, Risk Factors, Management, and Outcomes of Arterial and Venous Thromboembolic Complications) multicenter registry included patients with COVID‐19 from March 2020 to February 2024. A composite of venous and arterial thrombotic events and a composite of adjudicated cardiovascular events were assessed at 90 days after COVID‐19 diagnosis. Time‐to‐first‐event analyses were stratified by inpatients and outpatients and adjusted for competing risks. Outcomes were adjudicated by independent physicians. Results Of 11 165 patients, 4451 were inpatients (mean age, 65 years; 47% women) and 462 of 4451 were admitted to the intensive care unit. Among inpatients, the 90‐day cumulative incidences of thrombotic and cardiovascular events were 8.9% (95% CI, 8.0%–9.8%) and 15.6% (95% CI, 14.5%–16.7%), respectively. Corresponding estimates for patients in the intensive care unit were 20.0% (95% CI, 16.6%–24.0%) and 30.7% (95% CI, 26.8%–35.3%), while event rates were low among outpatients. There was an initial decline in cardiovascular events (adjusted hazard ratio [HR], 0.84; P=0.001) after the first COVID‐19 wave, with a subsequent catch‐up, with no significant linear change in cardiovascular events (adjusted Ptrend=0.634). Vaccination was associated with a reduced risk of cardiovascular events (adjusted HR, 0.75 [95% CI, 0.62–0.92]; P=0.005). Conclusions Thrombotic and cardiovascular events are common in hospitalized patients with COVID‐19, especially those in the intensive care unit, without a persistent decline over time. COVID‐19 vaccination was associated with a reduced risk of cardiovascular events.

Predicting wind speed is crucial in the wind energy sector, especially for controlling and balancing production and consumption in smart grids. However, wind speed is characterized by high uncertainty and variability. These fluctuations in wind speed are obvious in short term timescales. Deterministic wind speed forecasts, also named point predictions, do not consider the inherent uncertainties in wind speed predictions. Estimating these uncertainties is essential for providing reliable information to energy operators, enabling them to develop effective operational strategies. In this work, we introduce a probabilistic prediction intervals approach for short-term wind speed forecasting. The proposed framework consists of a hybrid model that integrates machine learning models and an automatic feature selection technique to estimate both point predictions and prediction intervals for wind speed. The simulation works demonstrate the effectiveness of our designed framework and demonstrate its ability to generate satisfactory prediction intervals in the most adopted evaluation criteria.

Background: An outbreak of an unknown respiratory illness caused by a novel coronavirus, SARS-CoV-2, emerged in the city of Wuhan in Hubei Province, China, in December 2019 and was referred to as coronavirus disease-2019 (COVID-19). Soon after, it was declared as a global pandemic by the World Health Organization (WHO) in March 2020. SARS-CoV-2 mainly infects the respiratory tract with different outcomes ranging from asymptomatic infection to severe critical illness leading to death. Different SARS-CoV-2 variants are emerging of which three have raised concerns worldwide due to their high transmissibility among populations. Objective: To study the prevalence of COVID-19 in the region of Nabatieh-South Lebanon during the past year and assess the presence of SARS-CoV-2 variants and their effect on the spread of infection during times of lockdown. Methods: In our study, 37,474 nasopharyngeal swab samples were collected and analyzed for the detection of SARS-CoV-2 virus in suspected patients attending a tertiary health care center in South Lebanon during the period between 16 March 2020 and 21 February 2021. Results: The results demonstrated a variation in the prevalence rates ranging from less than 1% during full lockdown of the country to 8.4% upon easing lockdown restrictions and reaching 27.5% after the holidays and 2021 New Year celebrations. Interestingly, a new variant(s) appeared starting January 2021 with a significant positive association between the prevalence of positive tests and the percentage of the variant(s). Conclusions: Our results indicate that the lockdown implemented by the Lebanese officials was an effective intervention to contain COVID-19 spread. Our study also showed that lifting lockdown measures during the holidays, which allowed indoor crowded gatherings to occur, caused a surge in COVID-19 cases and rise in the mortality rates nationwide. More importantly, we confirmed the presence of a highly transmissible SARS-CoV-2 variant(s) circulating in the Lebanese community from at least January 2021 onwards.

Background Sex impacts individuals' response to vaccination. However, most vaccine studies do not report these differences disaggregated by sex. The aim of this study was to assess sex differences in the immunogenicity and efficacy of influenza vaccine. Methods We performed a meta-analysis using phase 3 randomized controlled trial data conducted between 2010 and 2018. Using hemagglutination inhibition antibody titers for each strain, differences in geometric mean ratios (GMRs) were calculated by sex. Risk ratios (RRs) comparing seroconversion proportions were pooled for females and males using random-effects models. Vaccine efficacy (VE) was assessed. Data were analyzed by age group (18-64 vs >= 65 years). Results A total of 33 092 healthy adults from 19 studies were included for immunogenicity analysis, and 6740 from 1 study for VE. Whereas no sex differences in immunogenicity were found in adults <65 years old, older females had a significantly greater chance to seroconvert compared to older males for all strains: RRH1N1 = 1.17 [95% confidence interval {CI}, 1.12-1.23]; RRH3N2 = 1.09 [95% CI, 1.05-1.14]; RRVictoria = 1.23 [95% CI, 1.14-1.31]; RRYamagata = 1.22 [95% CI, 1.14-1.30]. GMRs were also higher in older females for all strains compared to older males. VE in preventing laboratory-confirmed influenza was higher in older females compared to older males with VEs of 27.32% (95% CI, 1.15%-46.56%) and 6.06% (95% CI, -37.68% to 35.90%), respectively. Conclusions Our results suggest a higher immunogenicity and VE in females compared to males in older adults. These differences in immunogenicity and VE support the disaggregation of vaccine data by sex in clinical trials and observational studies. Clinical Trials Registration CRD42018112260.

Forecasting wind speed plays an increasingly essential role in the wind energy industry. However, wind speed is uncertain with high changeability and dependency on weather conditions. Variability of wind energy is directly influenced by the fluctuation and unpredictability of wind speed. Traditional wind speed prediction methods provide deterministic forecasting that fails to estimate the uncertainties associated with wind speed predictions. Modeling those uncertainties is important to provide reliable information when the uncertainty level increases. Models for estimating prediction intervals of wind speed do not differentiate between daytime and nighttime shifts, which can affect the performance of probabilistic wind speed forecasting. In this paper, we introduce a prediction framework for deterministic and probabilistic short-term wind speed forecasting. The designed framework incorporates independent machine learning (ML) models to estimate point and interval prediction of wind speed during the daytime and nighttime shifts, respectively. First, feature selection techniques were applied to maintain the most relevant parameters in the datasets of daytime and nighttime shifts, respectively. Second, support vector regressors (SVRs) were used to predict the wind speed 10 minutes ahead. After that, we incorporated the non-parametric kernel density estimation (KDE) method to statistically synthesize the wind speed prediction errors and estimate the prediction intervals (PI) with several confidence levels. The simulation results validated the effectiveness of our framework and demonstrated that it can generate prediction intervals that are satisfactory in all evaluation criteria. This verifies the validity and feasibility of the hypothesis of separating the daytime and nighttime data sets for these types of predictions.

Community-driven projects have characterized Helicobacter pylori (Hp) infection in Indigenous communities in the Northwest Territories (NT) and Yukon (YT), Canada. These projects address concerns about the frequent diagnosis of Hp infection among community members and its relation to gastric cancer deaths, perceived to occur with alarming frequency in this region. Projects included breath-test screening for Hp infection, gastroscopy with gastric biopsies, and treatment to eliminate Hp infection. Previous project results showed a high prevalence of stomach pathologies associated with increased cancer risk among Hp-positive participants at baseline. This analysis describes changes in precancerous gastric pathologies in project participants who had gastroscopy before baseline treatment during 2008-2013 and again in 2017. Throughout the study period, the same pathologist graded Hp density, active gastritis, chronic gastritis, atrophic gastritis, and intestinal metaplasia using the updated Sydney System. Of 310 participants from three communities with baseline pathology data, 69 had follow-up pathology data. Relative to baseline, the prevalence of Hp infection and precancerous gastric pathology was substantially lower at follow-up; most participants who were Hp-positive at baseline and Hp-negative at follow-up had reduced severity of active, chronic, and/or atrophic gastritis at follow-up. Though follow-up numbers are small, these results yield evidence that successful Hp treatment has the potential to reduce the risk of gastric cancer in Arctic Indigenous communities.

Wind energy is one of the sustainable and clean energy resources that are uncertain because of its high fluctuation and stochastic volatility. Its uncertainty follows wind speed and occurs at multiple timescales. This fact necessitates the development of accurate wind speed prediction methods, especially for short-term periods. In this paper, we examine a two-step approach to improve the quality of ultra-short-term forecasting of wind speed using time series weather data and machine learning models. The Support Vector Regressor, Random Forests for regression, and Multi-layer perceptrons models for regression are the three machine learning models that are used in this work. The most relevant weather features are selected from original weather data using features selection methods and the predictions are obtained using separate machine-learning models. The simulation results identified the suitable combinations of feature selection techniques and machine learning models. They also showed that wrapper methods are the most suitable with support vector regressors and random forests in terms of performance and processing time.

The process of reusing ontologies is still challenging for the ontological community. One of the challenging efforts is to select the most relevant ontology from a set of candidates that needs a deep consideration. After the step of finding the candidates, many of them can be more appropriate than others as they fit better to the ontology requirements expressed by competency questions. First, we develop a mathematical formalisation based on Set Theory and we design the problem as an optimization problem to assist the knowledge engineer in selecting ontologies. Then, we provide formal steps to make well-founded comparison across a set of candidate ontologies. At last, we propose metrics to quantify the decision during the selection step.

Random Forest is an ensemble classification approach. It aims to design a discrete finite group of decision trees constructed based on bootstrap samples and random attribute selection. Random Forests have strong generalization capacities due to the variance in the training and attribute couple subsets used for constructing different decision trees in the forest. However, to construct a robust and effective random forest, two main issues need to be taken into account namely: (1) increasing the accuracy and diversity of decision trees; (2) decreasing the number of decision trees. In this paper, a genetic algorithm-based approach to tackle the aforementioned challenges related to random forest construction is proposed. Three objectives are taken into consideration. First, strengthening the classification accuracy of individual decision trees as well as that of the forest. Second, making use of diversity measures among the decision trees to improve the generalization of the constructed model. Third, minimizing the number of trees in the forest and finding an optimal subset of the random forest. An experimental evaluation on several datasets from the UCI Machine Learning Repository is conducted. The obtained results show that the proposed approach outperforms state-of-the-art classical as well as evolutionary random forest construction methods. Finally, the proposed approach is used to build a reliable random forest model for detecting Botnet traffic in Internet of Things environment.

Currently, numerous machine learning (ML) techniques are being applied in the field of renewable energy (RE). These techniques may not perform well if they do not have enough training data. Additionally, the main assumption in most of the ML algorithms is that the training and testing data are from the same feature space and have similar distributions. However, in many practical applications, this assumption is false. Recently, transfer learning (TL) has been introduced as a promising machine-learning framework to mitigate these issues by preparing extra-domain data so that knowledge may be transferred across domains. This learning technique improves performance and avoids the resource expensive collection and labeling of domain-centric datasets; furthermore, it saves computing resources that are needed for re-training new ML models from scratch. Lately, TL has drawn the attention of researchers in the field of RE in terms of forecasting and fault diagnosis tasks. Owing to the rapid progress of this technique, a comprehensive survey of the related advances in RE is needed to show the critical issues that have been solved and the challenges that remain unsolved. To the best of our knowledge, few or no comprehensive surveys have reviewed the applications of TL in the RE field, especially those pertaining to forecasting solar and wind power, load forecasting, and predicting failures in power systems. This survey fills this gap in RE classification and forecasting problems, and helps researchers and practitioners better understand the state of the art technology in the field while identifying areas for more focused study. In addition, this survey identifies the main issues and challenges of using TL for REs, and concludes with a discussion of future perspectives.

BACKGROUND Vaccine coverage for 18-month-old children in Canada is often below recommended levels, which may be partially due to parental forgetfulness. Text message reminders have been shown to potentially improve childhood immunization uptake, but were not widely utilized in Alberta, Canada. Additionally, it has been noted that language barriers may impede immunization service delivery but continue to be unaddressed in many existing reminder/recall systems. OBJECTIVE We aimed to assess the effectiveness and acceptability of using text messages, containing a link to online immunization information in different languages, to remind parents of their child’s 18-month immunization appointment. METHODS The Childhood Immunization Reminder Project (ChIRP) was a pilot intervention at two public health centres, one each in Lethbridge and Edmonton, Alberta, Canada. Two text message reminders were sent to parents: (1) a booking reminder 3 months before their child turned 18 months old, and (2) an appointment reminder three days before their scheduled appointment. Booking reminders included a link to the study website hosting immunization information in nine languages. To evaluate the intervention effectiveness, we compared absolute attendance no-show rates pre- and post-intervention. Acceptability of the intervention was evaluated through online surveys completed by parents and public health centre staff. Google Analytics was used to determine how often the online immunization information was accessed, from where, and in which languages. RESULTS Following the intervention, the Edmonton health centre had a reduction of 6.4% (95% CI: 3.0, 9.8) in appointment no-shows, with no change at the Lethbridge health centre (0.8%; 95% CI: -1.4, +3.0). Acceptability surveys were completed by 222 parents (response rate: 23.9%) and 22 staff. Almost all (>95%) respondents indicated that the reminders were helpful and had useful suggestions for improvements. All surveyed parents (100.0%) found it helpful to read the online immunization information in their language of choice. Google Analytics data showed that the immunization information was most often read in English (57.0%), Punjabi (25.1%), Arabic (6.3%), Spanish (5.8%), Italian (1.9%), Chinese (1.9%), French (1.0%), Tagalog (0.5%), and Vietnamese (0.5%). CONCLUSIONS The study findings support the use of text message reminders as a convenient and acceptable method to minimize parental forgetfulness and potentially reduce appointment no-shows. The diverse languages accessed in the online immunization information suggest the need to provide appropriate translated immunization information. Further research is needed to evaluate the impact of text message reminders on childhood immunization coverage in different settings.

Publicly funded immunization programs have grown in both complexity and scope, resulting in increased costs and more complex programmatic decision making. Economic evaluations can provide crucial information to support informed decision making. While very few countries have National Immunization Technical Advisory Groups that analyze economic information, many have started to develop processes for this purpose. Since these guidelines are being developed at the national level, we propose that regional jurisdictions, especially those responsible for healthcare (e.g., provinces, territories, states), need clear processes for incorporating this information into their immunization decision making and program implementation. We interviewed Canadian vaccine experts involved in provincial vaccine policy decision making to identify current practices, perceptions, and recommendations around incorporating economic analysis into that process. Based on these interviews, we make five recommendations: (1) economic evidence should be routinely incorporated into the decision making process; (2) economic experts should sit on, or be available to, regional advisory committees; (3) efforts should be made to build on regional expertise by increasing educational opportunities on economic evaluation; (4) processes should include guidelines for when economic analysis is not required; and (5) clarification on the role of regional advisory groups in economic analysis is needed in relation to national expertise. The information presented here provides a starting point for regional health policy experts and decision makers to work collaboratively with national partners to create transparent and effective approaches to incorporating economic analysis into vaccine decision making.

Objective To investigate COVID-19 vaccine uptake and intent among pregnant people in Canada, and determine associated factors. Methods We conducted a national cross-sectional survey among pregnant people from May 28 through June 7, 2021 ( n = 193). Respondents completed a questionnaire to determine COVID-19 vaccine acceptance (defined as either received or intend to receive a COVID-19 vaccine during pregnancy), factors associated with vaccine acceptance, and rationale for accepting/not accepting the vaccine. Results Of 193 respondents, 57.5% ( n = 111) reported COVID-19 vaccine acceptance. Among those who did not accept the vaccine, concern over vaccine safety was the most commonly cited reason (90.1%, n = 73), and 81.7% ( n = 67) disagreed with receiving a vaccine that had not been tested in pregnant people. Confidence in COVID-19 vaccine safety (aOR 16.72, 95% CI: 7.22, 42.39), Indigenous self-identification (aOR 11.59, 95% CI: 1.77, 117.18), and employment in an occupation at high risk for COVID-19 exposure excluding healthcare (aOR 4.76, 95% CI: 1.32, 18.60) were associated with vaccine acceptance. Perceived personal risk of COVID-19 disease was not associated with vaccine acceptance in the multivariate model. Conclusion Vaccine safety is a primary concern for this population. Safety information should be communicated to this population as it emerges, along with clear messaging on the benefits of vaccination, as disease risk is either poorly understood or poorly valued in this population.

Many renewable energy resources, including wind energy, are uncertain and often unavailable when needed, with high variability and dependency on atmospheric and climatic conditions. Variability and uncertainty of wind energy follow those wind speed and occur at multiple timescales; that is, from seconds to minutes and then to hours. They also require movement of other resources to ensure the balance between generation and load. This requires an accurate wind speed prediction. However, most existing wind speed forecasting models are based on data that do not take into account the difference between day and night, which can limit the accuracy of wind speed forecasting. Therefore, to improve the prediction accuracy, this study proposes three scenarios for wind speed prediction: first, the prediction model does not take into account the difference between day and night. The second model uses only data recorded during the day, and the third model deals with data measured during the night only. A comparative analysis and comprehensive evaluation will be given at the end of this work to verify the proposed hypothesis. The recorded climate data and measured wind speed at AUMET station will be used in this study.

Background Vaccine coverage for 18-month-old children in Canada is often below the recommended level, which may be partially because of parental forgetfulness. SMS text message reminders have been shown to potentially improve childhood immunization uptake but have not been widely used in Alberta, Canada. In addition, it has been noted that language barriers may impede immunization service delivery but continue to remain unaddressed in many existing reminder and recall systems. Objective This study aimed to assess the effectiveness and acceptability of using SMS text messages containing a link to web-based immunization information in different languages to remind parents of their child’s 18-month immunization appointment. Methods The Childhood Immunization Reminder Project was a pilot intervention at 2 public health centers, one each in Lethbridge and Edmonton, Alberta, Canada. Two SMS text message reminders were sent to parents: a booking reminder 3 months before their child turned 18 months old and an appointment reminder 3 days before their scheduled appointment. Booking reminders included a link to the study website hosting immunization information in 9 languages. To evaluate intervention effectiveness, we compared the absolute attendance no-show rates before the intervention and after the intervention. The acceptability of the intervention was evaluated through web-based surveys completed by parents and public health center staff. Google Analytics was used to determine how often web-based immunization information was accessed, from where, and in which languages. Results Following the intervention, the health center in Edmonton had a reduction of 6.4% (95% CI 3%-9.8%) in appointment no-shows, with no change at the Lethbridge Health Center (0.8%, 95% CI −1.4% to 3%). The acceptability surveys were completed by 222 parents (response rate: 23.9%) and 22 staff members. Almost all (>95%) respondents indicated that the reminders were helpful and provided useful suggestions for improvement. All surveyed parents (222/222, 100%) found it helpful to read web-based immunization information in their language of choice. Google Analytics data showed that immunization information was most often read in English (118/207, 57%), Punjabi (52/207, 25.1%), Arabic (13/207, 6.3%), Spanish (12/207, 5.8%), Italian (4/207, 1.9%), Chinese (4/207, 1.9%), French (2/207, 0.9%), Tagalog (1/207, 0.5%), and Vietnamese (1/207, 0.5%). Conclusions The study’s findings support the use of SMS text message reminders as a convenient and acceptable method to minimize parental forgetfulness and potentially reduce appointment no-shows. The diverse languages accessed in web-based immunization information suggest the need to provide appropriate translated immunization information. Further research is needed to evaluate the impact of SMS text message reminders on childhood immunization coverage in different settings.

Background Additional doses of COVID-19 vaccine have been proposed as solutions to waning immunity and decreased effectiveness of primary doses against infection with new SARS-CoV-2 variants. However, the effectiveness of additional vaccine doses relies on widespread population acceptance. We aimed to assess the acceptance of additional COVID-19 vaccine doses (third and annual doses) among Canadian adults and determine associated factors. Methods We conducted a national, cross-sectional online survey among Canadian adults from October 14 to November 12, 2021. Weighted multinomial logistic regression analyses were used to identify sociodemographic and health-related factors associated with third and annual dose acceptance and indecision, compared to refusal. We also assessed influences on vaccine decision-making, and preferences for future vaccine delivery. Results Of 6010 respondents, 70% reported they would accept a third dose, while 15.2% were undecided. For annual doses, 64% reported acceptance, while 17.5% were undecided. Factors associated with third dose acceptance and indecision were similar to those associated with annual dose acceptance and indecision. Previous COVID-19 vaccine receipt, no history of COVID-19 disease, intention to receive an influenza vaccine, and increasing age were strongly associated with both acceptance and indecision. Chronic illness was associated with higher odds of acceptance, while self-reported disability was associated with higher odds of being undecided. Higher education attainment and higher income were associated with higher odds of accepting additional doses. Minority first language was associated with being undecided about additional doses, while visible minority identity was associated with being undecided about a third dose and refusing an annual dose. All respondents reported government recommendations were an important influence on their decision-making and identified pharmacy-based delivery and drop-in appointments as desirable. Co-administration of COVID-19 and influenza vaccines was viewed positively by 75.5% of the dose 3 acceptance group, 12.3% of the undecided group, and 8.4% of the refusal group. Conclusions To increase acceptance, targeted interventions among visible minority and minority language populations, and those with a disability, are required. Offering vaccination at pharmacies and through drop-in appointments are important to facilitate uptake, while offering COVID-19/influenza vaccine co-administration may have little benefit among those undecided about additional doses.

Dear Editor: We thank Sookaromdee and Wiwanitkit (2022) for their letter to the Editor, and agree with the authors’ point that the influences on vaccine hesitancy and vaccine decisionmaking during pregnancy are multifaceted. However, we disagree with the emphasis they place on sociodemographic factors. Our finding of vaccine safety concerns as a primary driver of vaccine acceptance/hesitancy among pregnant people is well supported by the literature, as cited in our paper (Reifferscheid et al., 2022). Conversely, sociodemographic factors have not been found to be universally significant determinants of vaccine decisions (Nikpour et al., 2022), but are instead indicators of the context in which vaccine decisions are made. Even in settings where sociodemographic factors are significant, information on the risks and benefits of vaccination, tailored to the literacy and numeracy skills of pregnant individuals, is still required (Skirrow et al., 2022). We also think it is important to note that the vaccine decision-making process during pregnancy is unique, because the vaccine impacts both the pregnant individual and their unborn infant. Therefore, studies of the general population, such as those cited by Sookaromdee and Wiwanitkit (Khan et al., 2022; Xiao et al., 2022), should not be considered representative of the pregnant population. Sookaromdee and Wiwanitkit also note that our study was conducted at one specific point in time and that no firm conclusions should be made from an individual study. These are observations that we do not dispute. Rather than advocating for a static approach, the goal of our paper was to emphasize the importance of continuing to explore potential strategies to support pregnant people and their healthcare providers in vaccine decision-making.

The integration of solar energy in smart grids and other utilities is continuously increasing due to its economic and environmental benefits. However, the uncertainty of available solar energy creates challenges regarding the stability of the generated power the supply-demand balance’s consistency. An accurate global solar radiation (GSR) prediction model can ensure overall system reliability and power generation scheduling. This article describes a nonlinear hybrid model based on Long Short-Term Memory (LSTM) models and the Genetic Programming technique for short-term prediction of global solar radiation. The LSTMs are Recurrent Neural Network (RNN) models that are successfully used to predict time-series data. We use these models as base predictors of GSR using weather and solar radiation (SR) data. Genetic programming (GP) is an evolutionary heuristic computing technique that enables automatic search for complex solution formulas. We use the GP in a post-processing stage to combine the LSTM models’ outputs to find the best prediction of the GSR. We have examined two versions of the GP in the proposed model: a standard version and a boosted version that incorporates a local search technique. We have shown an improvement in terms of performance provided by the proposed hybrid model. We have compared its performance to stacking techniques based on machine learning for combination. The results show that the suggested method provides significant improvement in terms of performance and consistency.