Classification and Regression Tree (CART) Description

The Classification and Regression Tree (CART) method is yet another statistical method that looks to classify data depending on selected variables. Specialized software is required to use this method. Depending on the variables chosen, the endpoint is a flowchart, or decision tree (shown at right), that is then used to decide what a particular value might be, given basic conditions.

You should recall from the overview of the methods that the CART method requires a moderate level of effort to develop with a very low level of effort needed to use the method. Accuracy is pretty high. Using a decision tree is quite easy (one is presented on the next page), but, as in almost all of the methods, the user needs to be able to find and appropriately use various types of meteorology and air quality data. Forecast production time is very low (one hour or less), depending on the ability of the user in acquiring the required datatsets.