The aim of this project is to predict bookings that will be canceled.
It consists of two steps: preprocessing and modelling
Some metrics are as follows for files:
File | Algorithm | Accuracy (CV=10) | Precision | Recall | F1 Score | AUC |
---|---|---|---|---|---|---|
H1 | Random Forests | 0.8591 | 0.87 | 0.87 | 0.88 | 0.95 |
GBM | 0.8539 | 0.87 | 0.90 | 0.88 | 0.96 | |
XGB | 0.8533 | 0.86 | 0.90 | 0.88 | 0.95 | |
Light GBM | 0.8528 | 0.82 | 0.91 | 0.86 | 0.93 | |
H2 | Random Forests | 0.8545 | 0.90 | 0.73 | 0.80 | 0.94 |
GBM | 0.8815 | 0.88 | 0.83 | 0.85 | 0.95 | |
XGB | 0.8637 | 0.87 | 0.79 | 0.83 | 0.94 | |
Light GBM | 0.8111 | 0.86 | 0.65 | 0.74 | 0.90 |
The comparison between the prediction models as follows:
Random Forests | GBM | XGB | Light GBM | |
---|---|---|---|---|
H1 | ![]() |
![]() |
![]() |
![]() |
H2 | ![]() |
![]() |
![]() |
![]() |