Un marco de minería de datos para modelar las dinámicas de interacción entre las fincas y el entorno: modelado de rendimientos de cultivos en Ecuador.

Abstract

This work proposes a framework that addresses the limitations of current agricultural statis tics to develop yield prediction models. Currently, agricultural statistics at farm levels do not allow their integration and use together with market information, climatic variables and field management practices. This work, to date, represents a novel effort to solve this problem. This integration of information is crucial for understanding the dynamic interactions between production units and socio-ecosystems that impact agricultural yields. The thesis includes a study of data integration and its applications, including the construction of cross sectional databases, modeling of greenhouse gas emissions, and non-agricultural income prediction models. Additionally, a novel crop yield model was built, called a Crop Sequence Transformer, which integrates intermediate results to enhance the accuracy of yield estimation. Overall, this work contributes to the development of more comprehensive and accurate agricultural information systems, which can benefit farmers, policymakers, and other stakeholders in the agriculture industry.