Predicting Freeway Traffic Crash Severity Using XGBoost-Bayesian Network Model with Consideration of Features Interaction

Algorithm 1

K2 algorithm optimization process.

	Input: node variable set “n,” node order “order,” maximum number of parent nodes “Max_fan_in,” observation data set “D.”
	Output: Optimal network structure based on K2 algorithm, that is, each point and its parent node.
(1)	for i = 1 to n (n represents the number of total nodes);
(2)	π i = ∅;
(3)	p old =  (i, πi);
(4)	Set variable OKTOProceed = true;
(5)	while OKTOProceed = true and |πi| < m do;
(6)	pick one node z from pred (i) − πi, let (i, πi  {z}) maximize, where pred (i) denotes the nodes whose order is before the ith node;
(7)	Calculate ;
(8)	If then;
(9)	;
(10)	;
(11)	;
(12)	else ;
(13)	end while;
(14)	return (node , parent node );
(15)	end for;
(16)	end K2.