Interpreting user actions to better understand their needs provides an important tool for improving information access services. In the context of organic Web search, considerable effort has been made to model user behavior and infer query intent, with the goal of improving the overall user experience. Much less work has been done in the area of sponsored search, i.e., with respect to the advertisement links (ads) displayed on search result pages by many commercial search engines. This thesis develops and evaluates new models and methods required to interpret user browsing and click behavior and understand query intent in this very different context. The concern of the initial part of the thesis is on extending the query categories for commercial search and on inferring query intent, with a focus on two major tasks: i) enriching queries with contextual information obtained from search result pages returned for these queries, and ii) developing relatively simple methods for the reliable labeling of training data via crowdsourcing. A central idea of this thesis work is to study the impact of contextual factors (including query intent, ad placement, and page structure) on user behavior. Later, this information is incorporated into probabilistic models to evaluate the quality of advertisement links within the context that they are displayed in their history of appearance. In order to account for these factors, a number of query and location biases are proposed and formulated into a group of browsing and click models. To explore user intent and behavior and to evaluate the performance of the proposed models and methods, logs of query and click information provided for research purposes are used. Overall, query intent is found to have substantial impact on predictions of user click behavior in sponsored search. Predictions are further improved by considering ads in the context of the other ads displayed on a result page. The parameters of the browsing and click models are learned using an expectation maximization technique applied to click signals recorded in the logs. The initial motivation of the user to browse the ad list and their browsing persistence are found to be related to query intent and browsing/click behavior. Accommodating these biases along with the location bias in user models appear as effective contextual signals, improving the performance of the existing models.

京都大学 / 新制・課程博士 / 博士(工学) / 甲第23858号 / 工博第4945号 / 新制||工||1772(附属図書館) / 京都大学大学院工学研究科社会基盤工学専攻 / (主査)教授 KIM Chul-Woo, 教授 高橋 良和, 准教授 北根 安雄 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

[pt] Modelos estruturais para séries temporais vêm sendo bastante utilizados ultimamente e, adotam, basicamente, a mesma idéia da decomposição clássica de uma série temporal em seus componentes não-observáveis: tendência, sazonalidade, cíclica e irregular; para a componente cíclica, em particular, que é modelada por uma senóide a amortecida, existem apenas soluções no contexto da Estatística Clássica Harvey (1985). Neste trabalho discutimos extensivamente a solução Bayesiana para o modelo, tornando completamente estocástico a componente ciclo e obtendo um algoritmo para a estimação seqüencial dos parâmetros. A natureza não linear do problema é tratada pelos Modelos Dinâmicos Bayesianos; West e Harrison (1986). / [en] The structural models for time series, so much in use today make use of the well know idea of decomposing a time series into its unobserved components of trend, seasonal, cycle and noise. The cyclical component in particular, which uses a damped sine wave to describe its moviment, has a clear solution available already in computer packages on the Classica framework of Harvey (1985). In this thesis we present a Bayesian solution to the cyclical component modelled by the same damped sine wave. The frequency and the damping factor, regarded as hyperparameters on the Classical solution are now incorporated to the system state vector and estimated by a sequential procedure. Finally, the non-linear nature of model is elegantly dealt with by the Bayesian Dynamic Models of West and Harrison (1986).

[pt] Esta dissertação investiga a utilização do Modelo de Regressão Multivariada Seemingly Unrelated sob uma perspectiva Bayesiana, na estimação das curvas de carga dos principais eletrodomésticos. Será utilizada uma estrutura de Demanda Condicional (CDA), consideradas de especial interesse no setor comercial e residencial para o gerenciamento pelo lado da demanda (Demand Side Management) dos hábitos dos consumidores residenciais. O trabalho envolve três partes principais: uma apresentação das metodologias estatísticas clássicas usadas para estimar as curvas de cargas; um estudo sobre Modelos de Regressão Multivariada Seemingly Unrelated usando uma aproximação Bayesiana. E por último o desenvolvimento do modelo num estudo de caso. Na apresentação das metodologias clássicas fez-se um levantamento preliminar da estrutura CDA para casos univariados usando Regressão Múltipla, e multivariada usando Regressão Multivariada Seemingly Unrelated, onde o desempenho desta estrutura depende da estrutura de correlação entre os erros de consumo horário durante um dia específico; assim como as metodologias usadas para estimar as curvas de cargas. No estudo sobre Modelos de Regressão Multivariada Seemingly Unrelated a partir da abordagem Bayesiana considerou-se um fator importante no desempenho da metodologia de estimação, a saber: informação a priori. No desenvolvimento do modelo, foram estimadas as curvas de cargas dos principais eletrodomésticos numa abordagem Bayesiana mostrando o desempenho da metodologia na captura de ambos tipos de informação: estimativas de engenharia e estimativas CDA. Os resultados obtidos avaliados pelo método acima comprovaram superioridade na explicação de dados em relação aos modelos clássicos. / [en] The present dissertation investigates the use of multivariate regression models from a Bayesian point of view. These models were used to estimate the electric load behavior of household end uses. A conditional demand structure was used considering its application to the demand management of the residential and commercial consumers. This work is divided in three main parts: a description of the classical statistical methodologies used for the electric load prediction, a study of the multivariate regression models using a Bayesian approach and a further development of the model applied to a case study. A preliminary revision of the CDA structure was done for univariate cases using multiple regression. A similar revision was done for other cases using multivariate regression (Seemingly Unrelated). In those cases, the behavior of the structure depends on the correlation between a minimization of the daily demand errors and the methodologies used for the electric load prediction. The study on multivariate regression models (Seemingly Unrelated) was done from a Bayesian point of view. This kind of study is very important for the prediction methodology. When developing the model, the electric load curves of the main household appliances were predicted using a Bayesian approach. This fact showed the performance of the metodology on the capture of two types of information: Engineering prediction and CDA prediction. The results obtained using the above method, for describing the data, were better than the classical models.

[pt] Esta tese concentra-se essencialmente em modelos para análise e previsão de séries temporais vetoriais onde o comportamento sazonal é o foco principal, e no procedimento Bayesiano correspondente de estimação seqüencial. A estrutura básica do modelo multivariado proposto, não- linear, de crescimento sazonal multiplicativo para séries temporais, consiste de uma componente de tendência localmente linear para cada série individual e uma única componente sazonal, multiplicativa, compartilhada por todas as séries marginais. O procedimento de estimação seqüencial baseia-se em aproximações analíticas que viabilizam uma análise conjugada, representando uma extensão não-linear do algoritmo apresentado em Barbosa and Harrison (1992) para modelos lineares dinâmicos multivariados. Detalhes do modelo proposto e sua implementação são apresentados, assim como exemplos da aplicação do método, com dados simulados e reais. Para os dados reais, escolheu-se os valores do consumo de energia elétrica no Brasil, cuja metodologia de previsão adotada pelas empresas de energia também faz parte deste trabalho. / [en] This thesis is essentially devoted to models for analysis and forecasting of vector time series, where the seasonal behavior is the main focus, and a Bayesian procedure of sequential estimation is adopted. The basic structure of the non-linear multivariate model, of seasonal growth multiplicative for time series, consists of a locally linear trend component for each individual series and a shared multiplicative seasonal component common to all marginal series. The procedure of sequential estimation is based on analytic transformations to obtain a conjugate analysis, representing a non-linear extension of the algorithm by Barbosa and Harrison (1992) for multivariate dynamic linear models. Details of the proposed procedure and of the implementation are shown, as well examples of the application of the method, with simulated and real data. For real data, the brazilian electricity demand values were chose. The forecasting methodology adopted by the energy companies is also present in this work.

abstract: Functional traits research has improved our understanding of how plants respond to their environments, identifying key trade-offs among traits. These studies primarily rely on correlative methods to infer trade-offs and often overlook traits that are difficult to measure (e.g., root traits, tissue senescence rates), limiting their predictive ability under novel conditions. I aimed to address these limitations and develop a better understanding of the trait space occupied by trees by integrating data and process models, spanning leaves to whole-trees, via modern statistical and computational methods. My first research chapter (Chapter 2) simultaneously fits a photosynthesis model to measurements of fluorescence and photosynthetic response curves, improving estimates of mesophyll conductance (gm) and other photosynthetic traits. I assessed how gm varies across environmental gradients and relates to other photosynthetic traits for 4 woody species in Arizona. I found that gm was lower at high aridity sites, varied little within a site, and is an important trait for obtaining accurate estimates of photosynthesis and related traits under dry conditions. Chapter 3 evaluates the importance of functional traits for whole-tree performance by fitting an individual-based model of tree growth and mortality to millions of measurements of tree heights and diameters to assess the theoretical trait space (TTS) of “healthy” North American trees. The TTS contained complicated, multi-variate structure indicative of potential trade-offs leading to successful growth. In Chapter 4, I applied an environmental filter (light stress) to the TTS, leading to simulated stand-level mortality rates up to 50%. Tree-level mortality was explained by 6 of the 32 traits explored, with the most important being radiation-use efficiency. The multidimentional space comprising these 6 traits differed in volume and location between trees that survived and died, indicating that selective mortality alters the TTS. / Dissertation/Thesis / Doctoral Dissertation Biology 2017

This thesis studies the relationship between GDP and Social Progress Index, components of social progress model and their dimensions. Using the dataset of 49 countries and Bayesian Model Averaging (BMA) and clustering analysis we found that there is not straight relationship between GDP and SPI. By testing 15 different models for each of 3 dimension (Basic Human Needs, Foundations of Wellbeing and Opportunity) of SPI we have found that the best variation of components would be to include all of them for each dimension. By using BMA approach we have found that the best model of SPI out of 12 components includes only intercept, tolerance and inclusion variables. The rest of components show quite low probability of inclusion, however, none of them showed 0 posterior probability. JEL Classification A13, C11, E01, I30, Keywords Kuznets, progress, SPI, GDP, BMA Author's e-mail valeria.e.efimenko@gmail.com Supervisor's e-mail daniel.vach@gmail.com

We propose the use of Latent Class Analysis methods to analyze the covariate inclusion patterns across specifications resulting from Bayesian model averaging exercises. Using Dirichlet Process clustering, we are able to identify and describe dependency structures among variables in terms of inclusion in the specifications that compose the model space. We apply the method to two datasets of potential determinants of economic growth. Clustering the posterior covariate inclusion structure of the model space formed by linear regression models reveals interesting patterns of complementarity and substitutability across economic growth determinants.

Better understanding and prediction of tree growth is important because of the many ecosystem services provided by forests and the uncertainty surrounding how forests will respond to anthropogenic climate change. With the ultimate goal of improving models of forest dynamics, here we construct a statistical model that combines complementary data sources, tree-ring and forest inventory data. A Bayesian hierarchical model was used to gain inference on the effects of many factors on tree growth-individual tree size, climate, biophysical conditions, stand-level competitive environment, tree-level canopy status, and forest management treatments-using both diameter at breast height (dbh) and tree-ring data. The model consists of two multiple regression models, one each for the two data sources, linked via a constant of proportionality between coefficients that are found in parallel in the two regressions. This model was applied to a data set of similar to 130 increment cores and similar to 500 repeat measurements of dbh at a single site in the Jemez Mountains of north-central New Mexico, USA. The tree-ring data serve as the only source of information on how annual growth responds to climate variation, whereas both data types inform non-climatic effects on growth. Inferences from the model included positive effects on growth of seasonal precipitation, wetness index, and height ratio, and negative effects of dbh, seasonal temperature, southerly aspect and radiation, and plot basal area. Climatic effects inferred by the model were confirmed by a den-droclimatic analysis. Combining the two data sources substantially reduced uncertainty about non-climate fixed effects on radial increments. This demonstrates that forest inventory data measured on many trees, combined with tree-ring data developed for a small number of trees, can be used to quantify and parse multiple influences on absolute tree growth. We highlight the kinds of research questions that can be addressed by combining the high-resolution information on climate effects contained in tree rings with the rich tree-and stand-level information found in forest inventories, including projection of tree growth under future climate scenarios, carbon accounting, and investigation of management actions aimed at increasing forest resilience.

The Sahel region located in Western Africa is well known for its high rainfall variability. Severe and recurring droughts have plagued the region during the last three decades of the 20th century, while heavy precipitation events (with return periods of up to 1,200 years) were reported between 2007 and 2014. Vulnerability to extreme events is partly due to the fact that people are not prepared to cope with them. It would be of great benefit to farmers if information about the magnitudes of precipitation and streamflow in the upcoming rainy season were available a few months before; they could then switch to more adapted crops and farm management systems if required. Such information would also be useful for other sectors of the economy, such as hydropower production, domestic/industrial water consumption, fishing and navigation. A logical solution to the above problem would be seasonal rainfall and streamflow forecasting, which would allow to generate knowledge about the upcoming rainy season based on information available before it's beginning. The research in this thesis sought to improve seasonal rainfall and streamflow forecasting in the Sahel by developing statistical rainfall and streamflow seasonal forecasting models. Sea surface temperature (SST) were used as pools of predictor. The developed method allowed for a systematic search of the best period to calculate the predictor before it was used to predict average rainfall or streamflow over the upcoming rainy season. Eight statistical models consisted of various statistical methods including linear and polynomial regressions were developed in this study. Two main approaches for seasonal streamflow forecasting were developed here: 1) A two steps streamflow forecasting approach (called the indirect method) which first linked the average SST over a period prior to the date of forecast to average rainfall amount in the upcoming rainy season using the eight statistical models, then linked the rainfall amount to streamflow using a rainfall-runoff model (Soil and Water Assessment Tool (SWAT)). In this approach, the forecasted rainfall was disaggregated to daily time step using a simple approach (the fragment method) before being fed into SWAT. 2) A one step streamflow forecasting approach (called as the direct method) which linked the average SST over a period prior to the date of forecast to the average streamflow in the upcoming rainy season using the eight statistical models. To decrease the uncertainty due to model selection, Bayesian Model Averaging (BMA) was also applied. This method is able to explore the possibility of combining all available potential predictors (instead of selecting one based on an arbitrary criterion). The BMA is also capability to produce the probability density of the forecast which allows end-users to visualize the density of expected value and assess the level of uncertainty of the generated forecast. Finally, the economic value of forecast system was estimated using a simple economic approach (the cost/loss ratio method). Each developed method was evaluated using three well known model efficiency criteria: the Nash-Sutcliffe coefficient (Ef), the coefficient of determination (R2) and the Hit score (H). The proposed models showed equivalent or better rainfall forecasting skills than most research conducted in the Sahel region. The linear model driven by the Pacific SST produced the best rainfall forecasts (Ef = 0.82, R2 = 0.83, and H = 82%) at a lead time of up to 12 months. The rainfall forecasting model based on polynomial regression and forced by the Atlantic ocean SST can be used using a lead time of up to 5 months and had a slightly lower performance (Ef = 0.80, R2 = 0.81, and H = 82%). Despite the fact that the natural relationship between rainfall and SST is nonlinear, this study found that good results can be achieved using linear models. For streamflow forecasting, the direct method using polynomial regression performed slightly better than the indirect method (Ef = 0.74, R2 = 0.76, and H = 84% for the direct method; Ef = 0.70, R2 = 0.69, and H = 77% for the indirect method). The direct method was driven by the Pacific SST and had five months lead time. The indirect method was driven by the Atlantic SST and had six months lead time. No significant difference was found in terms of performance between BMA and the linear regression models based on a single predictor for streamflow forecasting. However, BMA was able to provide a probabilistic forecast that accounts for model selection uncertainty, while the linear regression model had a longer lead time. The economic value of forecasts developed using the direct and indirect methods were estimated using the cost/loss ratio method. It was found that the direct method had a better value than the indirect method. The value of the forecast declined with higher return periods for all methods. Results also showed that for the particular watershed under investigation, the direct method provided a better information for flood protection. This research has demonstrated the possibility of decent seasonal streamflow forecasting in the Sirba watershed, using the tropical Pacific and Atlantic SSTs as predictors.The findings of this study can be used to improve the performance of seasonal streamflow forecasting in the Sahel. A package implementing the statistical models developed in this study was developed so that end users can apply them for seasonal rainfall or streamflow forecasting in any region they are interested in, and using any predictor they may want to try.