Vehicle Driving Risk Prediction Based on Markov Chain Model

Table 6

MNL training results based on test set.

Variables	Future state: 1	Future state: 2	Future state: 3
	Initial state: (1) low risk
Driving mode	−0.474	−0.497	ref.
RL_avg	−0.569	−0.151	ref.
RL_last	−0.150	0.543	ref.
CON	0.153	−0.213	ref.
Constant	9.608	6.194	ref.
		−321.583
		−128.311
Rho-square		0.601
	Initial state: (2) medium risk
Driving mode	−0.105	−0.018	ref.
RL_avg	0.857	0.279	ref.
RL_last	1.587	0.599	ref.
CON	−0.672	−0.172	ref.
Constant	−0.421	2.483	ref.
		−144.884
		−26.989
Rho-square		0.814
	Initial state: (3) high risk
Driving mode	0.038	0.096	ref.
RL_avg	−0.025	−0.052	ref.
RL_last	−0.264	−0.619	ref.
CON	0.054	0.125	ref.
Constant	−27.884	−2.213	ref.
		−161.964
		−19.525
Rho-square		0.880

Note. “ref.” represents the reference state in MNL regression.   -test significance, .   -test significance, . “L(0)” represents initial likelihood of the model with constants only; “” represents likelihood of the specified model. Pseudo-.