The combination of neural estimates in prediction and decision problems

In this dissertation, different ways of combining neural predictive models or neural-based forecasts are discussed. The proposed approaches consider mostly Gaussian radial basis function networks, which can be efficiently identified and estimated through recursive/adaptive methods. Two different ways of combining are explored to get a final estimate – model mixing and model synthesis –, with the aim of obtaining improvements both in terms of efficiency and effectiveness. In the context of model mixing, the usual framework for linearly combining estimates from different models is extended, to deal with the case where the forecast errors from those models are correlated. In the context of model synthesis, and to address the problems raised by heavily nonstationary time series, we propose hybrid dynamic models for more advanced time series forecasting, composed of a dynamic trend regressive model (or, even, a dynamic harmonic regressive model), and a Gaussian radial basis function network. Additionally, using the model mixing procedure, two approaches for decision-making from forecasting models are discussed and compared: either inferring decisions from combined predictive estimates, or combining prescriptive solutions derived from different forecasting models. Finally, the application of some of the models and methods proposed previously is illustrated with two case studies, based on time series from finance and from tourism.

BlueFriends: measuring, analyzing and preventing social exclusion between elementary school students

Social exclusion is a relatively recent term, whose creation is attributed to René Lenoir(Lenoir, 1974). Its concept covers a remarkably wide range of social and economic problems, and can be triggered for various reasons: mentally and physically handicapped, abused children, delinquents, multi-problem households, asocial people, and other social “misfits” (Silver, 1995, pp. 63; Foucault, 1992). With an increasingly multi-cultural population, cultural and social inequalities rapidly ascend, bringing with them the need for educational restructuring. We are living in an evermore diverse world, and children need to be educated to be receptive to the different types of people around them, especially considering social and cultural aspects. It is with these goals that inclusive education has seen an increased trend in today’s academic environment, reminding us that even though children may be taught under the same roof, discriminatory practices might still happen. There are, however, a number of developed tools to assess the various dimensions of social networks. These are mostly based on questionnaires and interviews, which tend to be fastidious and don’t allow for longitudinal, large scale measurement. This thesis introduces BlueFriends, a Bluetooth-based measurement tool for social inclusion/exclusion on elementary school classes. The main goals behind the development of this tool were a) understanding how exclusion manifests in students’ behaviors, and b) motivating pro-social behaviors on children through the use of a persuasive technology. BlueFriends is a distributed application, comprised by an application running on several smartphones, a web-hosted database and a computer providing a visual representation of the data collected on a TV screen, attempting to influence children behaviors. The application makes use of the Bluetooth device present on each phone to continuously sample the RSSI (Received Signal Strength Indication) from other phones, storing the data locally on each phone. All of the stored data is collected, processed and then inserted into the database at the end of each day. At the beginning of each recess, children are reminded of how their behaviors affect others with the help of a visual display, which consists of interactions between dogs. This display illustrates every child’s best friends, as well as which colleagues they don’t interact with as much. Several tips encouraging social interaction and inclusiveness are displayed, inspiring children to change their behaviors towards the colleagues they spend less time with. This thesis documents the process of designing, deploying and analyzing the results of two field studies. On the first study, we assess how the current developed tools are inferior to our measuring tool by deploying a measurement only study, aimed at perceiving how much information can be obtained by the BlueFriends application and attempting to understand how exclusion manifests itself in the school environment. On the second study, we pile on the previous to try and motivate pro-social behaviors on students, with the use of visual cues and recommendations. Ultimately, we confirm that our measurement tool’s results were satisfying towards measuring and changing children’s behaviors, and conclude with our thoughts on possible future work, suggesting a number of possible extensions and improvements.