Scikit-learn

Note

• sklearn model.fit(X) normally accept df[['var1']] with double bracket, not single bracket for X

• PCA(n_components: int | float) float n_compoents can be percentage of variance to be explained

• sklearn model.transform normally output np.ndarray, not pd.DataFrame

• tn, fp, fn, tp = metrics.confusion_matrix(y_test, y_pred).ravel()

• fpr, tpr, thresholds = roc_curve(y_test, y_scores[:,1])

• Scikit-learn in Python: 100+ Data Science Exercises ex.57: k-means from scratch

• Scikit-learn in Python: 100+ Data Science Exercises ex.68: pca from scratch

• check GridSearch on udemy Scikit-learn in Python: 100+ Data Science Exercises ex.50

  # python code about GridSearch
  X_train, X_test, y_train, y_test = train_test_split(data, target)
  
  classifier = DecisionTreeClassifier()
  
  params = {'max_depth': np.arange(1, 10),
               'min_samples_leaf': [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20]}
  
  grid_search = GridSearchCV(
      classifier,
      param_grid=params,
      scoring='accuracy',
      cv=5
  )
  grid_search.fit(X_train, y_train)
  print(grid_search.best_params_)

Common Model

• train_test_split(X, y, test_size=0.33, stratify=y | None)
• GridSearch(model, param_grid: dict, scoring: str, cv:int)
  • clf.best_params_
• RandomForestClassifier
• LinearRegression
  • coeffs_
• LogisticRegression
• CountVectorizer(stop_words: list | Literal['english']) remove words