Innovation diffusion modelling
Since new products and technologies directly affect many aspects of life of people, communities, countries and economies, the diffusion of innovations appears a very important field of research. Modelling and forecasting the diffusion of innovations is a broad field of study, whose relevance is confirmed by the considerable body of research devoted to it. This research has especially called the attention on the importance of innovations in triggering the evolution of social and economic systems and on the central role played by diffusion models from a strategic and anticipative point of view. With more than 4000 publications since the 1940, it has been said that “no other field of behavioural science research represents more effort by more scholars in more disciplines in more nations” (Rogers, 2003).The phenomenon of innovation diffusion is essentially a social one, but has always attracted many researchers for its interdisciplinar character, allowing to combine theories and concepts from many disciplines, such as natural sciences, mathematics, physics, statistics, social sciences, marketing science, economics, technological forecasting and technology management. The formal representation of diffusion processes has historically used epidemic models borrowed from biology, namely the logistic or s-shape equation, under the hypothesis that an innovation spreads in a social system through communication between persons like an epidemy through contagion. The most famous and employed evolution of this equation is the Bass model (Bass, 1969), developed in the field of quantitative marketing and soon become a major reference, due to its surprisingly simplified formal structure on one hand and its predictive power on the other. Since the publication of the Bass model in Management Science in 1969, the field of quantitative marketing has become particularly strong in defining the boundaries and the directions of innovation diffusion theorizing and modelling. One of the characterizing aspects of the Bass model is that it addresses markets in aggregate: using aggregate adoption data it depicts and predicts the development of an innovation life cycle already in progress. In strategic terms, crucial forecasts concern the point of maximum growth of the life cycle, the peak, and the point of market saturation. Innovation diffusion models can be used both in a predictive way and for post hoc explanations, helping understand the evolution of a particular market, its response to various factors, like marketing strategies, incentive mechanisms, change in prices, policy measures. A timely investigation on the evolving structure of markets for innovations seems particularly crucial from an economic and managerial perspective, especially due to the shortening of innovation life cycles, the increasing level of competition between firms and products, the rise of successive generations of products. Some lines of research in this domain are currently evolving in order to account for complex network dynamics, network externalities, competition, forecasting with little or no data, and to develop new applications in several important fields, such as Energy, ICTs, Pharmaceuticals and Epidemiology, Service technologies.
Spatial dependence models for real estate markets: data integration and model building
A traditional research question in the statistical literature has to do with the dynamics of the real estate markets and the factors affecting them. Typically, house prices have been studied as function of house features, such as size, position and type of building. In this project, we aim at extending this stream of research by adding the effect of exogenous variables, that may exert a significant impact on price dynamics, namely the level of crime, the presence of Airbnb business, the distance from a metro station, the presence of schools within a given neighborhood. The ability to evaluate the effect of these variables is crucial for both business and policy perspectives. In doing this, we will consider the case of two cities, New York and Melbourne, for which this information is fully available as Open Data; dealing with Open Data gives rise to an interesting data integration problem, and the phase of data set construction will represent an essential insight of the project.
Because the data sets considered have a spatial cross-section nature, we need to account for spatial dependence between observations; we choose to employ two different modelling approaches incorporating the two, namely the Spatial Autoregressive Model, SAR, and the Spatial Error Model, SEM. An extremely important aspect in model building has to do with the variable selection process. We plan to perform model selection by extending the Stochastic Search Variable Selection (SSVS) algorithm. Specifically we propose a SSVS algorithm based on dirac spike and slab Lasso prior specifically tailored to select relevant covariates in the spatial regression context here considered.
Data Ethics
The project aims at investigating how an ethical dimension can be built into the process of business analytics. Business analytics is playing an increasing role due the growing availability of data onbehaviour and characteristics of individuals, and powerful data mining methods allowing to extract meaningful information from these. For example, social media users can take advantage of many services for free, and in return, companies get access to their data and elaborate them for business purposes, such as customer profiling. These actions are often performed thorugh algorthmic decision making. However, the potential harm -intended or unintened- arising from algorithmic decision making indicates that an ethical dimension is needed. The well known case of Cambridge Analytica has become a paradigmatic example of the problem. Although limitations and restrictions to the use and storage of personal data have been set by the recent GDPR, a business analytics process should go beyond the ‘codified rules’ and consider other principles, such as utility, rights, production of a common good. In this project, we aim to build a model for a business analytics process that explicitly takes into account the ethical dimension. In doing so, we plan to develop an ‘ethical framework’ to be applied to different business case studies. The literature on the topic is still very limited.