Vera Miguéis

Abstract
In the increasingly competitive fashion retail industry, companies are constantly adopting strategies focused on adjusting the products characteristics to closely satisfy customers' requirements and preferences. Although the lifecycles of fashion products are very short, the definition of inventory and purchasing strategies can be supported by the large amounts of historical data which are collected and stored in companies' databases. This study explores the use of a deep learning approach to forecast sales in fashion industry, predicting the sales of new individual products in future seasons. This study aims to support a fashion retail company in its purchasing operations and consequently the dataset under analysis is a real dataset provided by this company. The models were developed considering a wide and diverse set of variables, namely products' physical characteristics and the opinion of domain experts. Furthermore, this study compares the sales predictions obtained with the deep learning approach with those obtained with a set of shallow techniques, i.e. Decision Trees, Random Forest, Support Vector Regression, Artificial Neural Networks and Linear Regression. The model employing deep learning was found to have good performance to predict sales in fashion retail market, however for part of the evaluation metrics considered, it does not perform significantly better than some of the shallow techniques, namely Random Forest.

Abstract
Customers' response is an important topic in direct marketing. This study proposes a data mining response model supported by random forests to support the definition of target customers for banking campaigns. Class imbalance is a typical problem in telemarketing that can affect the performance of the data mining techniques. This study also contributes to the literature by exploring the use of class imbalance methods in the banking context. The performance of an undersampling method (the EasyEnsemble algorithm) is compared with that of an oversampling method (the Synthetic Minority Oversampling Technique) in order to determine the most appropriate specification. The importance of the attribute features included in the response model is also explored. In particular, discriminative performance was enhanced by the inclusion of demographic information, contact details and socio-economic features. Random forests, supported by an undersampling algorithm, presented very high prediction performance, outperforming the other techniques explored.

Abstract
Understanding the impact of corruption in modern societies, namely in standard of living, health and education services, is an issue that has attracted increased attention in recent years. This paper examines the relationship between the Corruption Perception Index (CPI) provided by Transparency International and the Human Development Index (HDI) of the United Nations Development Program and its components. The analysis is done for clusters of countries with similar levels of development. For the countries with high levels of development, it was found a negative relationship between corruption and human development. Moreover, for these countries, higher corruption levels are related to poor health care services, poor education services and low standard of living. For the other clusters of countries, these relationships were not statistically significant. The results obtained reinforce the importance of efforts by international politicians and organizations in fighting corruption, particularly in highly developed countries, to promote development. © Springer International Publishing AG 2017.

Abstract
This paper develops a decision support tool that can help fishery authorities to forecast bivalve landings for the dredge fleet accounting for several contextual conditions. These include weather conditions, phytotoxins episodes, stock-biomass indicators per species and tourism levels. Vessel characteristics and fishing effort are also taken into account for the estimation of landings. The relationship between these factors and monthly quantities landed per vessel is explored using multiple linear regression models and data mining techniques (random forests, support vector machines and neural networks). The models are specified for different regions in the Portugal mainland (Northwest, Southwest and South) using six years of data 2010-2015). Results showed that the impact of the contextual factors varies between regions and also depends on the vessels target species. The data mining techniques, namely the random forests, proved to be a robust decision support tool in this context, outperforming the predictive performance of the most popular technique used in this context, i.e. linear regression.

Abstract