Comparison of forecasting methodologies using egg price as a test case

Egg price forecasting of shelled eggs is a complex problem. Traditionally, future egg price has been predicted using a combination of regression analysis and experienced-based intuition to build a model, which is then fine-tuned to prevalent market conditions. Even after collecting reliable and expensive data, the subsequent analysis, in many cases, does not produce a high confidence to explain the variations in egg price. In the current project, a different approach using neural networks was used to forecast egg price. A neural network is a mathematical model of an information-processing structure that is loosely based on our present understanding of the working of human brain. An artificial neural network consists of a large number of simple processing elements connected to each other in a network. Urner Barry egg quotes from 1991 to 2002 as well as number of hens, egg storage capacity, and number of eggs placed for hatching from the USDA databases (1993 to 2000) were used to forecast egg price. Regression analysis explained only 37% of the variation in egg price due to the above-mentioned 3 factors. Neural networks, on the other hand, recognize the pattern in previous years' egg prices and then predict the future price more efficiently. The 3 networks used in this research (Ward, back-propagation, and general regression neural networks) fit the forecast line more tightly to the previous year's egg price line than did regression analysis. In the case of general regression neural networks, the R2 value was as high as 60%. Results suggest that neural networks may be a more reliable method of egg price forecasting than simple regression analysis if reliable data are collected and manipulated for such models.