Value of Image-based Yield Prediction: Multi-location Newsvendor Analysis
- 199 Downloads
Consider an agricultural processing company, which wants to pre-purchase crop from different locations before a harvesting season in order to maximize the total expected profit from all outputs subject to multiple resource constraints. The yields for different outputs are random and depend on the location. By remotely sensing from satellites or locally sensing from unmanned aerial vehicle, the firm may employ an image-based yield prediction model at the pre-purchase time. The distribution of the yield differs by a location. With the sensed data, the company updates the distribution of the yield using a regression model, whose explanatory variable is a vegetation index from image processing. At a more favorable location, the distribution of the yield is stochastically larger. The objective of this paper is to quantify the added value of image sensing in predicting crop yield. Specifically, the posterior yield distribution from image processing is used as an input to the multi-location newsvendor model with random yields. The optimal expected profit given the posterior distribution is compared to that with only the prior distribution of the yield. The difference between the total expected profits with the prior and posterior distributions is defined as the value of the sample information. We derive the type-1 and -2 errors as a function of the standard error of the estimate. In the numerical example, we show that the value of the sample information tends to be increasing (with diminishing return) as the yield prediction model becomes more accurate.
KeywordsImage processing Agricultural supply chain Applied operations research Stochastic model applications
The problem was materialized after some discussions with Mr. Chatbodin Sritrakul, our part-time master student who owns a rice mill in the Northeast of Thailand. His independent project, a part of requirement for a master’s degree in logistics management at the school, was related to our model.
- 3.Amaruchkul, K.: Newsvendor model for multi-inputs and -outputs with random yield: applciations to agricultural processing industries. In: Proceedings of the 8th International Conference on Operations Research and Enterprise Systems (ICORES 2019), Prague, Czech Republic, January 2019 (2019)Google Scholar
- 9.DeGroot, M., Schervish, M.: Probability and Statistics. Addison-Wesley, Boston (2002)Google Scholar
- 23.Niel, T., McVicar, T.: Remote sensing of rice-based irrigated agriculture: a review. Rice CRC Technical report R1105–01/01 (2001)Google Scholar
- 32.Turken, N., Tan, Y., Vakharia, A., Wang, L., Wang, R., Yenipazarli, A.: The multi-product newsvendor problem: review, extensions, and directions for future research. In: Choi, T. (ed.) Handbook of Newsvendor Problems. International Series in Operations Research & Management Science, vol. 176, pp. 3–39. Springer, New York (2012). https://doi.org/10.1007/978-1-4614-3600-3_1CrossRefGoogle Scholar
- 34.Wilasinee, S., Imran, A., Athapol, N.: Optimization of rice supply chain in Thailand: a case study of two rice mills. In: Sumi, A., Fukushi, K., Honda, R., Hassan, K. (eds.) Sustainability in Food and Water: An Asian perspective, vol. 18, pp. 263–280. Springer, Heidelberg (2010). https://doi.org/10.1007/978-90-481-9914-3_27CrossRefGoogle Scholar