Trees making
To build the tree we need to find answers to the following questions:
• Should all variables be simultaneously incorporated in building the tree?
• Does the accuracy of the tree depend on the number of records or variables?
• What combination of variables makes the best tree?
• Does the selection of target variables have significance in building the tree?
• Which data mining algorithm will reach the best conclusion?
For answering above questions, the used process for trees making are as follows:
• Step 1: Considering four identified groups are variables and two target variables, including several algorithms to build the decision tree, all applicable models will be evaluated. The existence of four independent variables, two target variables, and six feasible algorithms to build the decision tree (QUEST, CHAID, C5.0, CART in Towing state and CART in Ordered state) will result in 180 alternative ways to build the tree: