A Machine Learning Approach to Evaluate the Performance of Rural Bank

Algorithm 1

Gradient boosting regression tree with an adaptively reduced step size.

	Input:
	Training samples
	Residual tree training times is, random sampling rate is, complexity parameter is.
	Training steps:
	Initialize training samples, where, reduce step size,
	FOR j = 1, 2, …, M
(1)	Fromwithout replacement, repeat the subsample with a random ratio of rate as the training sample of the current regression tree.
(2)	Based on the complexity parameter, train theresidual tree modelon the current training sample.
(3)	Update reduction step .
(4)	Give the predicted valueof the training sampleon.
(5)	Update the output variable valueon the training sample.
	END FOR
	Output: improved gradient boosting regression tree model .