Table 51 Features by supervised ranking feature importance
From: IoT information theft prediction using ensemble feature selection
| XGBoost | CatBoost | Light GBM | RF |
|---|---|---|---|
| dport | sbytes | flgs_number | dport |
| bytes | dport | dport | state_number |
| dbytes | bytes | dur | bytes |
| rate | max | bytes | drate |
| dpkts | drate | state_number | srate |
| drate | dbytes | proto_number | dpkts |
| stddev | flgs_number | drate | dbytes |
| proto_number | stddev | rate | flgs_number |
| sum | min | stddev | proto_number |
| flgs_number | dpkts | sport | sum |
| dur | dur | dpkts | spkts |
| state_number | pkts | max | |
| mean | mean | dur | |
| spkts | dbytes | mean | |
| pkts | min | sbytes | |
| sport | max | min | |
| max | sum | pkts | |
| srate | sport | ||
| spkts | rate | ||
| sbytes |