Price modelling of agricultural commodities under consideration of external factors

Agricultural Commodities as Alternative Investment Products

With growing uncertainties in traditional financial products, commodities are becoming increasingly important as alternative investment products in a company's investment portfolio, for diversification purposes among others. An adequate and market-oriented modeling of price developments is therefore essential in order to derive important measures and key figures for the profitability of the investment, for risk management etc.

In this research project we create a link between appropriate price models for agricultural products and external factors which are of great importance for the price development. By means of predictions and projections of these factors price scenarios can be generated, which are used for example for the calculation of key figures in risk management (value-at-risk etc.).

Consideration of Commodity Specific Price Behaviour in the Two-Factor Model

As a starting point, we consider a two-factor model, which is based on classical financial mathematical approaches from share price and interest rate modelling. The model is able to depict the following price characteristics of agricultural products:

  • Seasonality
  • Long-term trends
  • Short-term fluctuations with return of mean value


The model is based on price time series of standardized futures products traded on the Chicago Board of Trade (CBoT).

In addition, time series of external factors that have a strong influence on price formation and development are taken into account for price modeling - such as:

  • Weather influences (precipitation, average temperature)
  • Demand and supply indicators (average consumption, production)
Preiszeitreihen von standardisierten Futures-Produkten
© Fraunhofer ITWM
Preiszeitreihen von standardisierten Futures-Produkten.

Data-Driven Price Scenarios Through the Use of Machine Learning Algorithms

We establish the connection between the two-factor model and time series of external factors by applying machine learning algorithms. In combination with the following methods:

  • Cluster analysis (K-Means, Hierarchical Agglomerative Clustering (HAC))
  • Classification (decision trees, neural networks)

The application of these simple, yet efficient algorithms ensures a simple model understanding and interpretability of the generated solutions. In practice, these solutions generated by machine learning offer added value in the decision support of a company.