An innovative machine learning workflow to research China’s systemic financial crisis with SHAP value and Shapley regression

Financial Innovation

Table 8 Interval effect

Features	Safe-haven zone	Notes
FinancialConcentration	[13, 35]	Concave impact; Compared with the feature value larger than the maximum of the safe-haven zone, the feature value less than the minimum of the safe-haven zone is more likely to increase the probability of risk
CPI-Dif	[3, 200(400)]
NarrowMoney	[7, 30]
NetExport	[2(0), 7]
DidYield	[0. 01, 0. 07]	Concave impact; Compared with the feature value less than the minimum of the safe-haven zone, the feature value larger than the maximum of the safe-haven zone is more likely to increase the probability of risk
GlobalNetExport	[− 3.5, 2]

The brackets 200(400) mean the same thing as above. The brackets 2(0) indicate that the safest range is [2,7], and the worst ratio cannot be less than 0