Model review and comparison
Predicting CTR with Machine Learning in Python
Kevin Huo
Instructor
Model review
from sklearn.linear_model import LogisticRegression
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.neural_network import MLPClassifier
- Logistic regression: linear classifier identifying decision boundary
- Decision trees: tree-format of conditions
- Random Forests: ensemble of Decision Trees
- Neural Networks (MLPs): layers using linear combinations of features with a nonlinear activation function
Model implementation
Similarities
- Feature transformation and regularization
- Fitting via
classifier.fit(X_train, y_train) - Predictions via
predict_proba()andpredict()
Differences
- Decision Trees:
max_depth,min_samples_split - Random Forests:
n_estimators,oob_score - Logistic Regression:
fit_intercept,class_weight - Neural Networks:
hidden_layer_sizes,max_iter
Model evaluation
- Key evaluation metrics:
- Confusion matrix:
confusion_matrix(y_test, y_pred) - Precision:
precision_score(y_test, y_pred) - Recall:
precision_score(y_test, y_pred) - F-beta score:
fbeta_score(y_test, y_pred, beta = 0.5) - AUC of ROC curve:
roc_auc_score(y_test, y_score[:, 1])
- Confusion matrix:
Main pros and cons of using neural networks
Pros
- Scalability with data
- Less need to do feature engineering
- More transferable across domains
Cons
- Less powerful on smaller datasets
- Difficult to interpret
- Computationally and financially cheaper
Let's practice!
Predicting CTR with Machine Learning in Python
