Qualitative Models of Climate Variations Impact on Crop Yields

Kotlowski, W. (2007). Qualitative Models of Climate Variations Impact on Crop Yields. IIASA Interim Report. IIASA, Laxenburg, Austria: IR-07-034

Preview

Text IR-07-034.pdf Download (370kB) | Preview

Abstract

This report presents application of machine learning methodology in modeling process. The objective is to identify and explain impact of weather variations on crop yields, and to test the approach on the agricultural and climatic data from the USA.

First, separation of weather and non-weather factors is performed by trend identification- two methods of trend identification are considered. Then the importance of the attributes is assessed using information gain measure and attributes are also aggregated into seasons. Finally, four types of classification methods (support vector machines, nearestneighbors classifier and two variants of decision rules induction) are applied to the data and the results are compared and analyzed.

The proposed approach differs from standard approaches to the crop yields modeling. It does not require a lot of expert knowledge, nor assume anything about the data distributions. All conclusions are drawn from data and the final model is build only from data. This approach is much simpler, however it maintains high accuracy and performance

Actions (login required)

View Item