Fig 1.
2015.12-2020.3 scale of internet users in China.
(a) 2015.12-2020.3 online shopping user scale and utilization rate. (b) 2015.12-2020.3 mobile online shopping user scale and utilization rate.
Fig 2.
The framework of the multi-graining scanning.
Fig 3.
The framework of the cascade forest structure.
Each cascade level consists of two random forests and two completely random forests.
Fig 4.
The framework of the improved deep forest.
Table 1.
The first five rows of item information data.
Table 2.
The first five rows of user information data.
Fig 5.
Daily user behavior data change graph.
Table 3.
Perdictive features of repurchase behavior of E—Commerce consumer.
Table 4.
Perdictive model parameter settings.
Table 5.
The cascade layer of improved deep forest parameters settings.
Fig 6.
Comparison results under the different number of features.
(a) Accuracy comparison chart. (b) AUC comparison chart. (c) FI-value comparison chart.
Table 6.
Performance comparison of predictive models.
Fig 7.
Performance comparison of predictive models.
Fig 8.
Comparison of training time between deep forest and CNN.