spam-classifier-randomforestclassifier

This is a simple machine learning model check emails weather it is spam or non-spam. I used tree classification models, linear models and ensemble models as well.

#import libraries for model development
from sklearn.tree import DecisionTreeClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.ensemble import AdaBoostClassifier, GradientBoostingClassifier, BaggingClassifier,RandomForestClassifier

Used machiene learning model accuracies

Note

Below model accuracies will be changed.

Ensemble Models

AdaBoostClassifier         - 0.9112221740038897    
GradientBoostingClassifier - 0.9257498754359741
BaggingClassifier          - 0.9203716026166482
RandomForestClassifier     - 0.9341813330761688

Tree Models

DecisionTreeClassifier - 0.8886299564427729

Linear Models

LogisticRegression - 0.8823480720703344

Hyperparameter-tuning

• The RandomForestClassifier model got the highest accuracy. Used it for further model development.

Used parameters :

param_grid = {'n_estimators' : [100, 200, 300],
              'criterion' : ['gini', 'entropy', 'log_loss'],
              'max_depth' : [10,20,30]}

Best Estimator :

RandomForestClassifier(criterion='entropy', max_depth=30, n_estimators=300)

Final accuracy : 0.963608

Dataset from UCI Machine Learning Repository Special Thanks for @dineshpiyasamara and @UCI