The strong emphasis placed on improving equality and well-being for all in the Sustainable Development Goals underscores the importance of tackling persistent within-country disparities in maternal mortality and poor health outcomes. Addressing maternal healthcare access barriers is, thus, crucial, particularly in low-resource settings. Numerous studies investigating determinants of maternal healthcare service use in Ethiopia exist but are limited by their focus on individual and household factors, and by methodological weaknesses. A nuanced understanding of the role of socioeconomic and geographic context in influencing access to care is needed to respond effectively. Maternity waiting homes (MWHs) are a potential strategy to address geographical barriers that delay women’s access to obstetric care. However, in addition to concerns about service quality, there is limited evidence on their effectiveness and on what models meet women’s needs. My research goals were, therefore, to contribute to the understanding of what contextual factors influence maternal healthcare service use in general; and to determine whether or not upgraded MWHs operating in an enabling environment could improve delivery care use in rural Ethiopia. My primary data sources were household surveys conducted as part of a cluster-randomized controlled trial evaluating MWHs and local leader training in Jimma Zone, Ethiopia. Random effects multivariable logistic regression analysis of survey data brought to light the social and financial resources that facilitate MWH use, highlighting the need for complementary interventions to make access more equitable. Spatial analyses identified subnational variation in service use at a finer scale than routinely reported and unmasked local variation in the relevance and magnitude of associations between individual-, interpersonal-, and health system factors and maternal healthcare use. These findings have implications for relying upon homogenous national responses to improve equality in access to care and health outcomes. Finally, analysis of trial data found a non-significant effect of interventions on delivery care use likely due to implementation issues and extraneous factors. The need to generate strong evidence of effectiveness of MWHs in improving maternal healthcare service use using sustainable and equitable MWH models using methods appropriate for complex intervention evaluation remains.

There are three ways to deal with component failure: reactive maintenance, preventive maintenance, and predictive maintenance. Reactive maintenance is to repair only once something breaks. Preventive maintenance is to repair before it breaks, independent of actual wear. Predictive maintenance is performed on the basis of real time operational data, repairing when components cross a certain degradation threshold.  With classification models one can determine the health state of a component. Regression models, on the other hand, allow the user to calculate a more precise estimate of remaining useful life. Previous research on regression models have exclusively used sensory data while classification models have used both sensory data as well as log-data. Research on predictive maintenance using regression models have found most success using SVM regression, decision trees, random forest regression, artificial neural networks and LSTM models.  Companies have more and more data to their disposal about the performance of their machines, but usually in the form of log-data. The goal of this research is to find if it is possible to use log-data for regression models. If this is the case, more sophisticated regression models can be used to apply predictive maintenance more accurately on a broader scale than is currently the case. The project was performed through a case study at a company in the semiconductor industry in the Netherlands, with years of log-data of their product that are gradually degrading over time. After quantifying the log-data and trying all kinds of different regression models in combination with different time scales, the results were unilaterally abysmal and were unable to make any decent prediction.  The reason for this according to several experts in the field of data science is that there was no in depth understanding of the data. They say it is required to have an integral understanding of the log-data and to closely collaborate with field engineers who know the data in and out. If a field engineer can say something about the degradation of a machine using only the log-data, a machine learning model can do it too. If a machine learning model is unable to purposefully overfit on the training data and the results are bad, there is no signal in the dataset and the task is impossible. It does not matter if the data was originally sensory or log-based, the only thing that matters is understanding what the data means and the presence of the degradation signal within.

No description available.

Turnouts are an important part of railway infrastructure for two reasons: infrastructure andmaintenance. For the infrastructure they provide the flexibility to allow the formulation and branchingof railway network and for maintenance they consume a large part of maintenance budget and have aprominent place in maintenance planning policy and activities. This is because as a “mechanical object”,a turnout often experiences malfunctions. The problem becomes even more complicated, since a turnoutis composed of many different parts and each of them fails for very different reasons (e.g. switch bladesvs crossing part). This is reflected in the different needs for maintenance activities, as railways areforced to pour in excessive amounts of resources to carry out emergency repairs, or to carry outunnecessary scheduled maintenance works in turnouts, which do not need to be inspected or repaired.Therefore, it is difficult to plan and organize maintenance activities in turnouts in an efficient manner.This raises the question of whether malfunctions in turnouts can be predicted and used as informationfor the maintenance planning process in order to optimize it and develop it into a more reliablepreventive maintenance planning.The aim of this analysis is to attempt to model the probability of various malfunctions in turnouts asa function of their main geometric and operational characteristics by using logistic regression modelsand then input these results into the maintenance planning process in order to optimize it. First, it wasimportant to objectify the railway track system and the turnout components, both in terms of parts andinterrelationships. Furthermore, the process and basic elements of railway maintenance planning weredefined, as well as arguments that motivate a turn towards preventive maintenance planningmethodologies. This was done through a comprehensive literature study.The basis of this research was case studies, which described the relationship between geometricaland operational characteristics of turnouts and their wear, as well as risk-based modelling methods inrailway maintenance planning. To create the analysis model, data from turnouts in eastern regionprovided by the Swedish Transport Administration were used, both from the point of view of describingthe underlying causes of turnout malfunctions and to formulate an object-oriented database suitable forusing in logistic regression models. The goal was a logit model that calculated the malfunctionprobability of a turnout, which could be used directly into a maintenance planning framework, whichranked maintenance activities in turnouts.The results obtained showed that although the model suffers from low correlation, differentrelationships between input variables and different functional errors were established. Furthermore, thepotential of these analytical models and modeling structures was shown to be able to developpreventive, predictive railway maintenance plans, but further analysis of the data structure is required,especially regarding data quality. Finally, further possible research areas are presented. / Spårväxlar är viktiga delar av järnvägens infrastruktur av två orsaker: infrastruktur och underhåll.För infrastrukturen ger de möjlighet till flexibla tillåter de formulering och grenning av järnvägsnät ochför underhållet konsumerar de en stor del av underhållsbudgeten och de har en framträdande plats iunderhållsplaneringspolitiken och aktiviteterna. Detta beror på att som ett ”maskinellt objekt”, harspårväxeln ofta fel. Problemet blir ännu mer komplicerat, eftersom en spårväxel består av många olikadelar och var och en av dem bryts ner av mycket olika skäl (t.ex. tunganordning vs korsningsdel). Dettaåterspeglas i olika behov av underhållsaktiviteter. Eftersom järnvägarna tvingas hålla alltför storamängder resurser för att utföra akuta reparationer eller för att utföra onödiga schemalagdaunderhållsarbeten i spårväxlar, som inte behöver inspekteras eller repareras. Därför är det svårt attplanera och organisera underhållsaktiviteter för spårväxlarna på ett effektivt sätt. Detta ställer fråganom funktionsfel i spårväxlar kan förutsägas och användas som information till  underhållsplaneringsprocessen för att optimera den och utveckla den till en pålitligare förebyggandeunderhållsplanering.Syftet med denna analys är att försöka modellera sannolikheten för olika funktionsfel i spårväxlarsom en funktion av deras huvudsakliga geometriska och operativa egenskaper med användning avlogistiska regressionsmodeller och sedan mata dessa resultat in i underhållsplaneringsprocessen för attoptimera den. För det första var det viktigt att objektifiera järnvägsspårsystemet ochspårväxlarkomponenterna, både vad gäller delar och inbördes förhållanden. Dessutom definieradesprocessen och grundelementen i järnvägsunderhållsplaneringen, samt att argument som motiverarförändring till förebyggande underhållsplaneringsmetoder. Detta gjordes genom en omfattandelitteraturstudie.Grunden i denna analys var fallstudier, som beskrev förhållandet mellan geometriska ochoperationella egenskaper hos spårväxlar och deras förslitning samt riskbaserade modelleringsmetoder ijärnvägsunderhållsplanering. För att skapa analysmodellen användes data från spårväxlar i östraregionen som tillhandahölls av Trafikverket, både ur synpunkten att beskriva de underliggandeorsakerna till spårväxlarsfel och för att formulera en objektorienterad databas lämplig för användning ilogistiska regressionsmodeller. Målet var en logitmodell som beräknade sannolikheten för fel i enspårväxel, som kunde användas direkt i en underhållsplaneringsram, som rangordnar lämpigaunderhållsaktiviteter i spårväxlar.Erhållna resultat visade att även om modellen lider av låg korrelation, konstaterades olika sambandmellan ingående variabler och olika funktionsfel. Vidare visades potentialen hos dessa analysmodelleroch modelleringsstrukturer för att kunna utveckla förebyggande, förutsägbarajärnvägsunderhållsplaner, men det krävs troligtvis ytterligare analys av datastrukturen, specielltangående datakvaliteten. Slutligen presenteras ytterligare möjliga forskningsområden.

In the last decades, the increasing number of new generation satellite images characterized by a better spectral, spatial and temporal resolution with respect to the past has provided unprecedented source of information for monitoring climate changes.To exploit this wealth of data, powerful and automatic methods to analyze remote sensing images need to be implemented. Accordingly, the objective of this thesis is to develop advanced methods for the analysis of multitemporal multispectral remote sensing images to support climate change applications. The thesis is divided into two main parts and provides four novel contributions to the state-of-the-art. In the first part of the thesis, we exploit multitemporal and multispectral remote sensing data for accurately monitoring two essential climate variables. The first contribution presents a method to improve the estimation of the glacier mass balance provided by physically-based models. Unlike most of the literature approaches, this method integrates together physically-based models, remote sensing data and in-situ measurements to achieve an accurate and comprehensive glacier mass balance estimation. The second contribution addresses the land cover mapping for monitoring climate change at high spatial resolution. Within this work, we developed two processing chains: one for the production of a recent (2019) static high resolution (10 m) land cover map at subcontinental scale, and the other for the production of a long-term record of regional high resolution (30 m) land cover maps. The second part of this thesis addresses the common challenges faced while performing the analysis of multitemporal multispectral remote sensing data. In this context, the third contribution deals with the multispectral images cloud occlusions problem. Differently from the literature, instead of performing computationally expensive cloud restoration techniques, we study the robustness of deep learning architectures such as Long Short Term Memory classifier to cloud cover. Finally, we address the problem of the large scale training set definition for multispectral data classification. To this aim, we propose an approach that leverages on available low resolution land cover maps and domain adaptation techniques to provide representative training sets at large scale. The proposed methods have been tested on Sentinel-2 and Landsat 5, 7, 8 multispectral images. Qualitative and quantitative experimental results confirm the effectiveness of the methods proposed in this thesis.

This study investigates the association between traditional and alternative working capital measures and the returns of industrial firms listed on the Johannesburg Stock E"change. Twenty five variables for all industrial firms listed for the most recent 10 years were derived from standardised annual balance sheet data of the University of Pretoria's Bureau of Financial Analysis. Traditional liquidity ratios measuring working capital position, activity and leverage, and alternative liquidity measures, were calculated for each of the 135 participating firms for the 1 0 years. These working capital measures were tested for association with five return measures for every firm over the same period. This was done by means of a chi-square test for association, followed by stepwise multiple regression undertaken to quantify the underlying structural relationships between the return measures and the working capital measures. The results of the tests indicated that the traditional working capital leverage measures, in particular, total current liabilities divided by funds flow, and to a lesser e"tent, long-term loan capital divided by net working capital, displayed the greatest associations, and e"plained the majority of the variance in the return measures. At-test, undertaken to analyse the size effect on the working capital measures employed by the participating firms, compared firms according to total assets. The results revealed significant differences between the means of the top quartile of firms and the bottom quartile, for eight of the 13 working capital measures included in the study. A nonparametric test was applied to evaluate the sector effect on the working capital measures employed by the participating firms. The rank scores indicated significant differences in the means across the sectors for si" of the 13 working capital measures. A decrease in the working capital leverage measures of current liabilities divided by funds flow, and long-term loan capital divided by net working capital, should signal an increase in returns, and vice versa. It is recommended that financial managers consider these findings when forecasting firm returns. / Business Management / D. Com. (Business Management)

Les modèles de mélange pour la régression sont utilisés pour modéliser la relation entre la réponse et les prédicteurs, pour des données issues de différentes sous-populations. Dans cette thèse, on étudie des prédicteurs de grande dimension et une réponse de grande dimension. Tout d’abord, on obtient une inégalité oracle ℓ1 satisfaite par l’estimateur du Lasso. On s’intéresse à cet estimateur pour ses propriétés de régularisation ℓ1. On propose aussi deux procédures pour pallier ce problème de classification en grande dimension. La première procédure utilise l’estimateur du maximum de vraisemblance pour estimer la densité conditionnelle inconnue, en se restreignant aux variables actives sélectionnées par un estimateur de type Lasso. La seconde procédure considère la sélection de variables et la réduction de rang pour diminuer la dimension. Pour chaque procédure, on obtient une inégalité oracle, qui explicite la pénalité nécessaire pour sélectionner un modèle proche de l’oracle. On étend ces procédures au cas des données fonctionnelles, où les prédicteurs et la réponse peuvent être des fonctions. Dans ce but, on utilise une approche par ondelettes. Pour chaque procédure, on fournit des algorithmes, et on applique et évalue nos méthodes sur des simulations et des données réelles. En particulier, on illustre la première méthode par des données de consommation électrique. / Finite mixture regression models are useful for modeling the relationship between a response and predictors, arising from different subpopulations. In this thesis, we focus on high-dimensional predictors and a high-dimensional response. First of all, we provide an ℓ1-oracle inequality satisfied by the Lasso estimator. We focus on this estimator for its ℓ1-regularization properties rather than for the variable selection procedure. We also propose two procedures to deal with this issue. The first procedure leads to estimate the unknown conditional mixture density by a maximum likelihood estimator, restricted to the relevant variables selected by an ℓ1-penalized maximum likelihood estimator. The second procedure considers jointly predictor selection and rank reduction for obtaining lower-dimensional approximations of parameters matrices. For each procedure, we get an oracle inequality, which derives the penalty shape of the criterion, depending on the complexity of the random model collection. We extend these procedures to the functional case, where predictors and responses are functions. For this purpose, we use a wavelet-based approach. For each situation, we provide algorithms, apply and evaluate our methods both on simulations and real datasets. In particular, we illustrate the first procedure on an electricity load consumption dataset.

A Euterpe edulis (palmeira juçara) é uma espécie de alta densidade na Floresta Atlântica e sofre processo de extinção devido ao desmatamento e ao corte ilegal para a extração do palmito. Uma alternativa para conservação é o manejo sustentável, utilizando seus frutos para a produção de polpa. O objetivo do trabalho é fornecer subsídios de mensuração para um sistema de manejo, visando à produção de frutos e polpa, através de modelos preditivos e, avaliação da estrutura, dinâmica e a regeneração natural das palmeiras. O levantamento foi realizado no Sertão do Ubatumirim, Ubatuba-SP, em uma área de bananal e uma de floresta secundária, durante 2011 e 2012. Para a construção dos modelos, foram selecionadas matrizes produtoras acompanhadas durante o período de frutificação. Foram coletadas medidas dendrométricas e os frutos maduros foram colhidos, pesados, despolpados e analisada sua massa seca. Foi avaliada qualitativamente a luminosidade recebida pelas matrizes. Foram ajustados e selecionados modelos para a predição da biomassa dos frutos e polpa seca. As variáveis preditoras foram: DAP, altura, e as indicadoras: área, ano e luz, com efeito simples e de interação. As variáveis resposta foram modeladas por regressão linear, com e sem transformação logarítmica. Os critérios de seleção dos modelos foram o coeficiente de determinação ajustado e o valor do Critério de Informação de Akaike (AIC). Para o levantamento das populações, em cada área foram locadas dez parcelas circulares, com 14m de raio, sendo mensuradas e identificadas as espécies arbóreas com DAP >= 5cm. Dentro das parcelas instituiu-se subparcelas para a amostragem da regeneração natural da juçara; com raio de 4,5m para os estádios de desenvolvimento das varas e arvoretas e raio de 3m para plântulas e mudas. Foi estimada a taxa de mortalidade de toda população e das palmeiras adultas. De cada estádio estimou-se os indivíduos por hectare e taxa de mudança anual. A regeneração encontrada foi comparada com uma estrutura padrão, sem ações antrópicas. Foi estimada a biomassa de frutos e de polpa da juçara nas áreas estudadas através de modelos lineares. Os modelos com transformação logarítmica apresentaram bons ajustes sendo os mais adequados os que utilizaram a combinação da variável indicadora altura ou variável combinada (diâmetro ao quadrado vezes altura). Os resultados mostraram efeito de interação da variável combinada e variável ano, o que indicou efeito nulo no ano de alta produtividade. Independente do tamanho das palmeiras, a produção foi igual, e efeito positivo crescente no ano de baixa produtividade. Os levantamentos mostraram que a área de formação secundária possui maior riqueza de espécies que a do bananal, com uma área basal duas vezes maior e menor taxa de mortalidade. Em ambas as áreas, os estádios plântulas e regeneração não se assemelham à população padrão, resultados estes mais drásticos na área do bananal. Isto pode levar à redução da população, com risco de desaparecimento. Na estimativa produtiva dos frutos por matriz, verificou-se o aumento entre os anos de produção. Quando se considera a produção por hectare, ocorre diminuição da produção na área do bananal devido a elevada mortalidade. / Euterpe edulis (juçara palm) occurs in hight density populations in the Atlantic Forest, but due the illegal exploration of its heart palm is a extinction threaten species. An alternative for its conservation is the sustainable management of its populations for fruit production. This study aims to provide measurement and biometric information for the development of a sustainable management system for fruit harvest. Prediction models for fruit and fruit pulp production were developed and the structure and natural regeneration of two populations were studied. The study was carry out in \"Sertão do Ubatumirim\", Ubatuba, in the state of São Paulo, where a banana plantation and a secondary forest were surveyed in the years 2011 and 2012. In order to develop the fruit prediction models, producing fruit palm trees were monitored throughout the period of fructification. Height and diameter measurements were taken in these trees and the ripe fruits were harvested, weighted, pulped and the fruit pulp dry weight was determined. The light intensity received by each tree was qualitatively determined by a ordinal scale with 5 levels. Prediction models for fruit biomass and fruit pulp dry weight were fitted by linear regression. Prediction quantitative variables were diameter (DBH) and height, while prediction qualitative variables, add to the model as indicator variables, were: forest type (banana plantation and secondary forest), harvest year (2011 and 2012) and light intensity. The criteria for selection of the models were adjusted coefficient of determination and the value of the Akaike Information Criteria (AIC). For the population surveys, circular plots of 14m radius were located in the study area, and all trees with DBH equal or greater than 5cm were measured and had its species identified. For the survey of juçara natural regeneration, circular subplots of 4.5m radius, concentric to adult tree plots, were established to enumerate small trees and saplings, and circular subplots of 3m radius were established to enumerate seedlings and small seedlings. Mortality rate was estimated of the entire population and adult juçara palms. For each plant development stage (tree, small tree, sapling, seedling and small seedling), the number of individuals and the annual change rate were also estimated. Best prediction models for individual palm tree fruit and fruit pulp production were logarithmic models, and prediction variable was tree height, followed by the combined variable (squared diameter times height). There was a clear interaction effect of the quantitative prediction variable (combined variable) and the qualitative prediction variable year, as indicator variable. In the hight production year, the quantitative prediction variable was not relevant for fruit and fruit pulp prediction, while in the low production year, there was a positive relationship between production and the quantitative prediction variable. Study sites were very different in forest structure and dynamics. As expected, the secondary forest site had higher species richness than the banana plantation site, as well lower mortality rate and twice its basal area. Compared to what is expected to sustainable juçara populations without human influence, both sites showed lower number of regenerating individuals (seedlings and small seedlings). Over the years, this fact, combined to the harvest of fruits for fruit pulp production, can represent risk to the sustainability of juçara populations in the study sites.

Neste trabalho são apresentadas aplicações estatísticas e extensões dos modelos Rathie-Swamee. Propostos em Rathie e Swamee (2006), os modelos Rathie-Swamee foram desenvolvidos a partir de uma generalização da distribuição logística. Esses modelos apresentam grande flexibilidade, assumindo formas unimodais e multimodais, e têm algumas aplicações exemplificadas neste trabalho com dados bimodais de pesca de camarões e de erupções de gêisers. Por meio de simulações desses modelos, são avaliados os desempenhos de diferentes métodos para obtenção de intervalos de confiança para os parâmetros dos modelos e dos estimadores de máxima verossimilhança. As extensões apresentadas para os modelos Rathie-Swamee são referentes à incorporação de covariáveis nos modelos, obtendo-se modelos de regressão. Esses novos modelos de regressão são utilizados para ajuste aos dados de pesca e de erupções, para exemplificar algumas aplicações dos modelos. Uma nova distribuição de probabilidades é apresentada como distribuição resultante de produtos e quocientes entre variáveis aleatórias independentes com distribuições Rathie-Swamee. Para essa nova distribuição é apresentada uma tabela com alguns quantis de interesse para diferentes valores do parâmetro, assim como os resultados de estimação por máxima verossimilhança obtidos para as simulações realizadas com diferentes valores para o parâmetro e tamanhos de amostra. / Applications and extensions to the Rathie-Swamee models are presented in this work. Proposed by Rathie and Swamee (2006), the Rathie-Swamee models were developed as a generalization to the logistic distribution. These models have great flexibility, assuming unimodal and multimodal shapes, and have some of its applications exemplified with bimodal data of shrimp fishing and geyser eruptions. By the use of simulations, the performance of different methods to obtain confidence intervals are compared. The extensions presented for the Rathie-Swamee models refer to the inclusion of covariates, creating regression models. These new regression models are fitted to fishing and eruption data, to exemplify some applications of the models. A new probability distribution is presented as the resulting distribution of quotients and products between independent random variables with Rathie-Swamee distributions. For this new distribution are presented some simulation results along with a table of quantiles for some percentage points of interest.

Building performance simulations [BPS] tools are important in all the design stages, mainly in the early ones. However, some barriers such as time, resources and expertise do not contribute to their implementation in architecture offices. This research aimed to develop regression models (meta-models) to assess the thermal discomfort in a Brazilian low-cost house [LCH] during early design. They predicted the degree-hours of discomfort by heat and/or by cold as function of the design parameters changes for three Brazilian cities: Curitiba/PR, São Paulo/SP, and Manaus/AM. This work focused on using the meta-models to evaluate the impact of the parameters related to natural ventilation strategies on thermal performance in LCH. The analyzed Brazilian LCH consisted in a naturally ventilated representative unit developed based on the collected data. The most influential parameters in thermal performance, namely as key design parameters, were building orientation, shading devices positions and sizes, thermal material properties of the walls and roof constructive systems as well as window-to-wall ratios (WWR) and effective window ventilation areas (EWVA). The methodology was divided into: (a) collecting projects of Brazilian LCH, and based on that a base model that was able to represent them was proposed, (b) defining the key design parameters and their ranges, in order to compose the design space to be considered, (c) simulating thermal performance using EnergyPlus coupled with a Monte Carlo framework to randomly sample the design space considered, (d) using the greater part of the simulation results to develop the meta-models, (e)using the remaining portion to validate them, and (f) applying the meta-models in a simple design configuration in order to test their potential as a support design tool. Overall, the meta-models showed R2 values higher than 0.95 for all climates. Except for the regression models to predict discomfort by heat for Curitiba (R2 =0.61) and São Paulo (R2 =0.74). In their application, the models showed consistent predictions for WWR variations, but unexpected patterns for EWVA. / Simulações do desempenho de edificações são ferramentas importantes em todo processo de desenvolvimento do projeto, especialmente nas etapas iniciais. No entanto, barreiras como tempo, custo e conhecimento especializado impedem a implementação de tais ferramentas nos escritórios de arquitetura. A presente pesquisa se propôs a desenvolver modelos de regressão (meta-modelos) para avaliar o desconforto térmico em uma habitação de interesse social [HIS] brasileira. Estes meta - modelos predizem os graus-hora de desconforto por calor ou por frio em função de alterações nos parâmetros de projeto para três cidades brasileiras: Curitiba/PR, São Paulo/SP e Manaus/AM. O foco deste trabalho é o uso dos meta-modelos para avaliar o impacto de parâmetros relacionados com estratégias de ventilação natural no conforto térmico em HIS. A HIS brasileira analisada consistiu em uma unidade representativa, naturalmente ventilada e desenvolvida baseada em dados coletados. Os parâmetros que mais influenciam o conforto térmico, nomeados parâmetroschave de projeto foram: orientação da edificação, posição e tamanho das proteções solares, propriedades térmicas dos sistemas construtivos das paredes e do telhado, assim como, áreas de janela nas fachadas e áreas efetiva de abertura. A metodologia foi dividida em: (a) coleta de projetos de HIS brasileiras que embasaram a proposição de um modelobase que os representassem, (b) definição dos parâmetros chave de projeto e suas faixas de variação, a fim de compor o universo de projeto a ser explorado, (c) simulações térmicas usando o EnergyPlus acoplado com uma ferramenta de Monte Carlo para variar randomicamente o universo de projeto considerado, (d) uso da maior parte dos resultados das simulações para o desenvolvimento dos meta-modelos,(e) uso da porção remanescente para a validação dos meta-modelos e (f) aplicação dos meta-modelos em uma simples configuração de projeto, visando testar o seu potencial como ferramenta de suporte de projeto. De modo geral, os meta-modelos apresentaram R2 superiores a 0,95 para todos os climas, exceto os meta-modelos para predizer desconforto por calor para Curitiba (R2 =0,61) e São Paulo (R2 =0,74). Na fase de aplicação, os modelos mostraram predições consistentes para variações na área de janela na fachada, mas incoerências para variações nas áreas efetiva de abertura.