Linmei Shang, Jifeng Wang, David Schäfer, Thomas Heckelei, Juergen Gall, Franziska Appel, Hugo Storm
{"title":"Surrogate modelling of a detailed farm-level model using deep learning","authors":"Linmei Shang, Jifeng Wang, David Schäfer, Thomas Heckelei, Juergen Gall, Franziska Appel, Hugo Storm","doi":"10.1111/1477-9552.12543","DOIUrl":null,"url":null,"abstract":"<p>Technological change co-determines agri-environmental performance and farm structural transformation. Meaningful impact assessment of related policies can be derived from farm-level models that are rich in technology details and environmental indicators, integrated with agent-based models capturing dynamic farm interaction. However, such integration faces considerable challenges affecting model development, debugging and computational demands in application. Surrogate modelling using deep learning techniques can facilitate such integration for simulations with broad regional coverage. We develop surrogates of the farm model FarmDyn using different architectures of neural networks. Our specifically designed evaluation metrics allow practitioners to assess trade-offs among model fit, inference time and data requirements. All tested neural networks achieve a high fit but differ substantially in inference time. The Multilayer Perceptron shows almost top performance in all criteria but saves strongly on inference time compared to a Bi-directional Long Short Term Memory.</p>","PeriodicalId":14994,"journal":{"name":"Journal of Agricultural Economics","volume":"75 1","pages":"235-260"},"PeriodicalIF":3.4000,"publicationDate":"2023-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1111/1477-9552.12543","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Agricultural Economics","FirstCategoryId":"96","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1111/1477-9552.12543","RegionNum":2,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"AGRICULTURAL ECONOMICS & POLICY","Score":null,"Total":0}
引用次数: 0
Abstract
Technological change co-determines agri-environmental performance and farm structural transformation. Meaningful impact assessment of related policies can be derived from farm-level models that are rich in technology details and environmental indicators, integrated with agent-based models capturing dynamic farm interaction. However, such integration faces considerable challenges affecting model development, debugging and computational demands in application. Surrogate modelling using deep learning techniques can facilitate such integration for simulations with broad regional coverage. We develop surrogates of the farm model FarmDyn using different architectures of neural networks. Our specifically designed evaluation metrics allow practitioners to assess trade-offs among model fit, inference time and data requirements. All tested neural networks achieve a high fit but differ substantially in inference time. The Multilayer Perceptron shows almost top performance in all criteria but saves strongly on inference time compared to a Bi-directional Long Short Term Memory.
期刊介绍:
Published on behalf of the Agricultural Economics Society, the Journal of Agricultural Economics is a leading international professional journal, providing a forum for research into agricultural economics and related disciplines such as statistics, marketing, business management, politics, history and sociology, and their application to issues in the agricultural, food, and related industries; rural communities, and the environment.
Each issue of the JAE contains articles, notes and book reviews as well as information relating to the Agricultural Economics Society. Published 3 times a year, it is received by members and institutional subscribers in 69 countries. With contributions from leading international scholars, the JAE is a leading citation for agricultural economics and policy. Published articles either deal with new developments in research and methods of analysis, or apply existing methods and techniques to new problems and situations which are of general interest to the Journal’s international readership.