摘要：Classification and Regression Trees (CART), a computer intensive nonparametric classification method, was used to model weekly Los Angeles wholesale prices (1990-93) for twelve different melon types. CART explained more of the variation in melon prices than did an ordinary least squares (OLS) regression with dummy variables. Explanatory variables ranked as the most-to-least important by CART are as follows: week, type of melon, year, size, grade, and shipping container. The most notable price change occurs when prices fall after 13 May.
