Master of Agribusiness / Department of Agricultural Economics / Vincent Amanor-Boadu / The objective of the research is to understand the volatility of cotton acres and estimate planted acres based on the factors that drive volatility in the United States at a regional level. Estimating cotton acres is important so that demand for cotton seed and technology can be anticipated and the appropriate investments in cotton seed production can be made.
Post Multi-Fiber Arrangement, the US cotton economy has entered a state of imperfect completion which makes cotton price, ending stocks and the relationship of cotton to other crops important in understanding volatility in cotton acres. Linear Regression, Random Forest and Partial Least Squares Neural Networks (PLS NN) were used to estimate cotton acres at a US and Regional Level. The modeling approaches used to estimate change in acres yielded similar performance for U.S. total, Southwest, and West. The PLS NN was slightly better for the Delta and Southeast, where more crop alternatives exist. Random Forest offered a different perspective on variable importance in all regions.
| Identifer | oai:union.ndltd.org:KSU/oai:krex.k-state.edu:2097/15534 |
| Date | January 1900 |
| Creators | Holmes, Beth |
| Publisher | Kansas State University |
| Source Sets | K-State Research Exchange |
| Language | en_US |
| Detected Language | English |
| Type | Thesis |
Page generated in 0.0018 seconds