# Chi-Squared significance test for stopping criteria in decision tree

Going through the paper of BFTree(Best First Decision Tree) from (Haijian Shi (2007). Best-first decision tree learning. Hamilton, NZ). I read for pre-pruning do a local attribute selection. And the goal is to find whether the attribute is significantly correlated with the class.(stop further splitting if no attributes are significantly correlated with the class otherwise do the splitting). For that matter chi-squared test is used as the Statistical significant test.

$\sum((obs_{freq}-exp_{freq})^2/exp_{freq})$