Hackathon( network activity anomaly detection)
No missing data was found. Used Smote to oversample the data as the dataset was imbalanced. Scaled the train, valid and test data using Standard Scaler. Dropped one feature named service and encoded the protocoltype and flag feature using label encoder. Then tried different algorithms to test my model. I used Random Forest classifier, Xg Boost , Decision Tree classifier, Support Vector Machine(SVM), K-nearest neighbors(KNN) and Adaboost. Out of these I got 100 % test accuracy on SVM and Random Forest followed by Xgboost and KNN at 99.98% accuracy & lastly had worst accuracy through adaboost at 79.06 % accuracy. I predicted the output of models and saved them by converting them into CSV file with "attack" as header.
Tags:
#python
#classification
#machine-learning