Datamining and big freight transport database

Nowadays, there is a large interest in developing a mathematical tool in the field of freight transport modeling for investigating the effects of investments and policies, involving large number of resources. However, freight demand forecasting models are still in the evolution stage [21] for the following reasons: • lower seniority (about 10 years) than the respective passenger models; • high number of decision-makers to consider (companies, shippers, carriers, logistics operators, port operators, deposits, etc.); • variety of products transported (in terms of categories, dimensions, weight, value, etc.); • high variability in decision-making processes; • limited availability of information (data often aggregated, dated, partial, heterogeneous, etc.). To take into account the complexity of freight transport system, researchers have proposed a wide array of models belonging to the aggregate or disaggregate model types [1] and to three different fields: the modelling of the relationship between transportation and economic activity, logistic decision making and processes and the link between traffic flows and networks [2].The objective is mainly to understand quantitative and qualitative aspects of future traffic demand and evaluate possible future scenarios according to most relevant and influencing variables of the freight market [7]. We also want to overcome the above-mentioned problems with a newfreight demand forecasting framework based on Bayesian Networks and using European official and available data. The model has to be easy to implement, not onerous and give probabilistic results in less time, with an estimation error similar to the more complex methods. It should be capable of giving the order of magnitude of forecasted freight flows for strategic decision making at a very early phase of policy development, and be complementary to more traditional, more precise, but much more expensive freight models for later stages of analysis.