Applying Classification Trees for Prediction of Sex Preference in Iran
|
Mahsa Saadati * , Arezoo Bagheri |
|
|
Abstract: (4700 Views) |
Employing a decision tree when classical methods are not applicable in the presence of various observations and variables becomes vital and important. Exploratory analysis, model detection, prediction and make decisions about large data are the most common usage of this method. If the response variable is categorical variable, it is known as classification tree. Comparing three different extraction methods, CART, CHAID, and QUEST of classification trees to predict sex preference for children of pre-marriage women by SPSS 22 was the main objective of this study. The data on this study consist of 6360 women referred to health centers to get premarital counseling which was collected by the multi-stage random sampling in 2014 in the whole country. The results showed that accuracy of all three classification tree algorithms in predicting sex preference for children are almost the same, but with regard to compliance CART to the existing theories, this model was considered as the final model. CART is a non-parametric method with easy interpretation feature that enables quick calculations and acquisitions to provide accurate results. According to the results of this tree, the ideal number of children, education level and women age are variables that affect sex preference.
|
|
Keywords: Decision tree, classification tree, CART algorithm, CHAID algorithm, QUEST algorithm. |
|
Full-Text [PDF 964 kb]
(1917 Downloads)
|
Type of Study: Applicable |
Subject:
Special Received: 2016/10/15 | Accepted: 2017/06/18 | Published: 2017/09/27
|
|
|
|
|
Send email to the article author |
|