The Challenge

Competing services or means to acquire soybean futures contracts price forecasts are expensive, require special financial knowledge or in-house analytic skills, and infrastructures.

Evidently, small and medium size traders and trading organizations are in need of systems supporting soy trading decisions. Futures contracts are agreements to buy or sell an asset at a future date at an agreed-upon price. Futures have standardized terms and are traded on an exchange, where prices are settled on a daily basis until the end of the contract. Companies and individuals can trade futures contracts either for hedging or speculative reasons. Therefore, predictability of soybeans futures contract price moves is very important to agricultural organizations, food companies and traders, that produce, purchase or do business with soybeans products through these contracts.

To put the challenge into context:

The worth of soybean derivatives import in EU is at 10 billion Euros per year.
The price volatility of Futures Contracts is in the range of 15%.

That accounts for 1.5 Billion Euros is potential losses!

SPATIAL’s Innovation

SPATIAL is developing two distinct Machine Learning (ML) models, one for soybean crop yield forecasting and one for prediction of soybeans futures contracts price moves.

They will be fed by EO data, including ERA5 Reanalysis Copernicus Climate Change Services (C3S), Landsat 7 and 8 data in the visible, near infrared and shortwave infrared parts of the E/M spectrum and Sentinel-2 data that offers additional bands in the vegetation Red Edge. Crop rotation practices together with the target to scale up globally after the end of the proof-of-concept have dictated the necessity to study in-season crop mapping techniques for which Sentinel-1 Synthetic Aperture Radar data are also necessary. The Earth Observation (EO) dataset covers the US region, being US the largest soybean producer globally.

The multi-parametric and multi-feature AI-system will be ultimately offered as a service in reasonable cost for soybeans commodities purchasing, trading and SME companies helping them fortify multi-million soybeans purchasing decisions against price risks.

A system supporting soy trading decisions for all

The machine learning models of SPATIAL offer to the users: 

  • In-season crop mapping
  • Soybean Yield Forecasting.
  • Soybean Futures Prices forecasting for multiple time horizons.

While the explainable AI allows for model feature analytics, to identify which feature has significant impact on the model’s results.  

The users may benefit from the SPATIAL’s features to:

  • Improve trading strategies
  • Enhance risk management
  • Increase trading profits

The Technology

SPATIAL interface will present the prediction, the prediction result and a series of economic and other metadata, so that the user can appreciate the quality and the success of the prediction, providing at the same time, useful information regarding the economic conditions at the time of the reference date. Finally, the system will present an overview on how the prediction was calculated, giving the contribution of each feature to support an Explainable AI paradigm.

In terms of performance, SPATIAL managed to forecast with less than 7% average relative error, the US soybean production, for the years 2016-2020. Its forecasting capabilities for soybeans futures contract prices traded in Chicago Board of Trade ranged from 63%, down to 60% for forecasts up to 3-months ahead of time.  

More precisely: 

  • 63% Accuracy for 2-weeks ahead forecasts
  • 63% Accuracy for 1-month ahead forecasts
  • 60% Accuracy for 2-months ahead forecasts
  • 60% Accuracy for 3-months ahead forecasts

Prime Contractor

HYPERTECH, in its long course has formed a dynamic portfolio of high-level services in the fields of digital strategy consulting and the development of IT products and services with an emphasis on the implementation of innovations that arise as a result of research and development. HYPERTECH’s activities and research interests focus on machine learning data analysis, financial markets and stock trading among others. In particular, HYPERTECH’s a-Quant division has already developed a range of Fintech products and solutions for professional investors that it has on the market today. Indicatively, the company provides, among others: portfolio management tools for brokers, institutional investors and managers of large funds, investment signal services for foreign exchange markets for professional investors (VIP-FX Signals, DSD4FX products) which are based on advanced machine learning algorithms (regime detection) , clustering, cointegration, Hidden Markov Models) to generate these signals, automatic multi-indicator technical analysis services in multiple time periods with the help of machine learning algorithms to extract signals with high predictive ability (T-Explorer product).

Sub-contractor

The National Observatory of Athens (NOA) is a Greek Research Centre active in Space Sciences and their applications with remarkable achievements. NOA with its three Institutes (Astronomy, Astrophysics, Space Applications & Remote Sensing IAASARS; Environmental Research and Sustainable Development IERSD; Geodynamics GI), its highly-skilled human resources and the important infrastructure obtained over the last two decades, plays an important role in international Space science activities and has a leading role in the national efforts for presence in the European Space Sector. NOA also constitutes a critical national link with the European Space Agency (ESA) and other relevant organizations and bodies within the European Union. IAASARS/NOA has been actively involved in Space Sciences, Space applications and Earth Observation with many achievements in leading research, and operational activities in the context of EU flagship programs/initiatives namely COPERNICUS, GALILEO, GEO, and GEOSS. The team involved in this proposal has extensive experience in the development of satellite-derived services for various applications.