A comprehensive method for medium-term analysis and forecasting (CMAF) of global monthly prices of agricultural commodities

This is a proposal for a comprehensive methodology for medium-term analysis and forecasting (CMAF) of monthly global prices of agricultural commodities.

The project will create a tool based on this methodology, which provides a detailed explanation of the forecasts and enables their complete interpretation.

Integrating eight econometric and machine learning (ML) methods, CMAF will combine the joint effects of over 100 possible variables. In addition, it will consider the inclusion of additional potential explanators for specific needs or purposes.

It will use different cross-validation techniques to avoid a priori research assumptions and realistically captures these complex relations.

First, the learning process begins with comprehensive stationary and causality tests, which detect the nature of each possible variable and its suitability to serve as an explanatory factor in the changed agricultural commodities prices.

Secondly, it performs a retrospective analysis while considering many variables from three different groups: market fundamentals, financial and climatic.

Thirdly, it uses relative importance analysis to reduce the number of features and include only those most essential for an accurate agricultural commodities price forecasting performance.

Lastly, it provides a detailed and intelligible visual interpretation of the results and the learning process in a straightforward manner to serve even those with no academic or financial background. CMAF will be easily trained using publicly available data and will be made available open source.

It will also be adaptable and could be applied to forecast the prices of various agricultural commodities, irrespective of budget, language or other skills constraints.

The outcome will be a powerful and broadly applicable tool that will promote understanding in the global food trade and thus enhance food security and social equity.