Proposal of new models for prediction of the cost of agricultural raw materials in a business intelligence and machine learning context

In this paper, we propose a data model for prediction of the cost of raw materials in a business intelligence context. Our contribution focuses initially on the implementation of a model with a star representation. This model highlights the fact (cost) to be predicted according to the axes linked to it. Secondly, from this basic model, our contribution is based on sub-models enabling us to carry out mono-dimensional anal-yses of the 'cost' fact. Thirdly, from these sub-models we establish associated mathematical models that allow us to deduce a global mathe-matical model from our basic model using linear regression and artificial neural network techniques. The implementation of these mono-dimensional sub-models in a machine learning database management system 'Minds DB', produces results that allow the prediction of raw material costs. Also, the predictions made by “Minds DB” are computationally validated by linear regression techniques which give better results than those of artificial neural networks.