Artificial Intelligence is hot, and at least one of your debates has probably touched on the fact that AI is not just glorified statistics. Also, putting together a PhD researcher with an industry researcher can easily turn a small talk into a passionate and intense conversation. Is it about Machine Learning versus Statistics? Industry research versus academic research? Is it about competition between the fields, or can they go hand in hand? If so, where is the benefit?
Here at MILNESIUM, we believe in collaboration. We like to see how academic statistical modelling and sub-fields of AI can help each other, resulting in their accelerated evolution. This was showcased by our colleague Florin Jurchis at the 13th International Conference on Applied Statistics, which took place in Bucharest in June 2019. The audience was able to observe a pragmatic collaboration between academia and industry by combining PhD research in Cybernetics and Statistics with our company’s internal Deep Learning Computer Vision R&D. As one of the results, it was once again emphasized how statistics can reduce AI bias and how sub-fields of AI can improve statistical modelling.
The main aim of the presented study was to analyze health status in the EU at regional NUTS2 level together with its influential socio-economic factors. Apart from statistics and classical econometrics, spatial analysis was conducted to determine possible similarities and dissimilarities among regions, taking into account the fact that the events of a specific region are interrelated with the events of neighboring regions. The negative distribution of the dependent variable, life expectancy, involved the Quantile Spatial Autoregressive Model, which was also applied to observe factor influence across different regions of the health-status proxy distribution.
Moreover, the analysis led to the anticipated conclusion that the greater the gaps between rich and poor, or between less and better educated groups, the greater the differences in health status and life expectancy. This highlighted the need for policies designed to reduce territorial health disparities across regions. In addition, Computer Vision and Deep Learning techniques were used to gather accurate data on urban green-space variables, given that more than half of the global population lives in urban areas and urban greenery has a strong positive influence on health.