Applications of Machine Learning Models in Regional and Demographic Economic Analysis: A Literature Survey

Big data is becoming increasingly available in regional and demographic economic analysis. Old challenges related to data unavailability are being replaced with difficulties dealing with more complex data sets that are often updated almost in real-time. Model-based, causality-oriented approaches are being substituted with black-box prediction-oriented techniques—primarily in nonacademic fields for expediting policy and management decisions. Big data sets allow for highly flexible relationships between variables, and machine learning (ML) techniques can enable researchers to discover highly complex relationships (Varian 2014). Complexity in this regard is related to “the unpredictable nature of non-linear and dynamic systems” (Nijkamp et al. 2001). Alpaydin (2016) defines ML as a way to achieve artificial intelligence (AI) and states that ML, which is grounded in statistical theory, is the driving force and a requirement for AI. The latter, in turn, can be defined as “The automation of activities that we associate with human thinking, activities such as decision-making, problem solving, and learning” (Bellman 1978). Regarding the ability of ML to discover complex relationships, Harding and Hersh (2018) state: