Development of Machine Learning Based Predictive Modeling for Forest Biomass Estimation**

基于机器学习的森林生物量估算预测模型的开发**

• 批准号: 537549-2018
• 负责人: Heppler, Glenn
• 金额: $ 1.82万
• 依托单位: University of Waterloo
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2018
• 资助国家: 加拿大
• 起止时间: 2018-01-01 至 2019-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=653464
• 关键词: Development; Machine; Learning; Based; Predictive

The objective of this research project is to develop efficient predictive modeling methods for forest biomass estimation applications as part of the industrial partner's (Hegyi Geomatics International Inc.) ongoing projects that deal with the quantification of forest biomass using satellite images and data analytics tools. Hegyi Geomatics develops decision support tools for natural resources management by integrating field observation, ecosystem models, Geographic Information System (GIS) and remote sensing. Toward this effort, Hegyi Geomatics is looking to develop a machine learning approach to analyze satellite images and estimate the forest biomass. Efficient predictive modeling approaches can be utilized to correlate band surface reflectance with forest biomass. To take full advantage of data-driven approaches, the specific problem investigated in this work is to develop a methodology to automatically select the most dominant inputs and to generate parsimonious predictive models using limited data sets for estimating forest biomass. The proposed predictive modeling methods will be developed utilizing advanced machine learning techniques by the research team from the University of Waterloo in close collaboration with the technical experts from the industrial partner. The benefits of the proposed predictive modeling methods include improved prediction accuracy, faster and more cost-effective predictive models, and better interpretations of constructed models. The proposed research effort to develop machine learning based predictive modeling methods for enabling effective and efficient data-driven modeling of complex systems also has significant implications for the GIS solutions and software that Hegyi Geomatics provides. Incorporation of the proposed predictive modeling methods into its data analytics and other decision support tools will help Hegyi Geomatics to further establish world-leading competitiveness of its services and products. The success of this project will enable the industrial partner to create new source of revenue generation and reach out to new clientele.

该研究项目的目标是为森林生物量估计应用开发有效的预测建模方法，作为产业合作伙伴(Hegyi Geomatics International Inc.)的一部分。正在进行的利用卫星图像和数据分析工具量化森林生物量的项目。Hegyi Geomatics通过集成野外观测、生态系统模型、地理信息系统(GIS)和遥感，开发用于自然资源管理的决策支持工具。为此，Hegyi Geomatics正在寻求开发一种机器学习方法来分析卫星图像并估计森林生物量。可以利用有效的预测建模方法将波段表面反射率与森林生物量进行关联。为了充分利用数据驱动的方法，这项工作研究的具体问题是开发一种方法来自动选择最主要的输入，并使用有限的数据集生成简约的预测模型来估计森林生物量。滑铁卢大学的研究小组将与工业合作伙伴的技术专家密切合作，利用先进的机器学习技术开发拟议的预测建模方法。所提出的预测建模方法的好处包括提高了预测精度、更快且更具成本效益的预测模型以及更好地解释所构建的模型。开发基于机器学习的预测建模方法以实现对复杂系统的有效和高效的数据驱动建模的拟议研究工作，对Hegyi Geomatics提供的地理信息系统解决方案和软件也具有重要影响。将建议的预测建模方法整合到其数据分析和其他决策支持工具中，将有助于Hegyi Geomatics进一步建立其服务和产品的世界领先竞争力。该项目的成功将使工业合作伙伴能够创造新的收入来源，并接触到新的客户。