Transparent Evaluation

Model Performance

Metrics below are generated from the held-out test split when the model is trained.

Accuracy99.52%
Precision99.7%
Recall99.18%
F1-Score99.44%

Training Information

ModelRandom Forest Classifier
DatasetUCI PhiUSIIL Phishing URL Dataset
Total URLs235370
Safe134850
Phishing100520
Train / Test188296 / 47074
Features32

Confusion Matrix

Pred SafePred PhishingActual Safe2690961Actual Phishing16519939

Top Feature Importance

uses_https43.115%
path_length13.339%
slash_count12.617%
special_count5.122%
digit_ratio4.593%
digit_count4.303%
url_length3.985%
entropy2.61%
domain_length2.331%
domain_digit_count1.951%
domain_parts1.771%
dot_count1.042%