多类别逻辑回归分类简介

1iris_df = pd.read_csv("Iris.csv")
2print(iris_df.head())

1iris_df.drop('Id', axis=1, inplace=True) # dropping Id column
2iris_df['target']=iris_df.Species.astype('category').cat.codes # creating codes for categories in target column
3# display taget values and their respective counts
4print(iris_df['target'].value_counts().to_string())

1X = iris_df.drop(["Species", "target"], axis=1)  # features or predictors
2y = iris_df.target  # target
3print(X[0:2], y[0:2])
4
5# Feature scaling
6from sklearn.preprocessing import StandardScaler, MinMaxScaler
7# scaler = StandardScaler()
8scaler = MinMaxScaler()
9X = scaler.fit_transform(X)
10print(X[0:2], y[0:2])

1from sklearn.linear_model import LogisticRegression
2# Creating model instances
3logR_ovr = LogisticRegression(multi_class='ovr') # one-vs-rest
4# fitting the model
5logR_ovr.fit(X,y)
6# Accuracy Score 
7print("Score when multi_class='ovr':",logR_ovr.score(X,y))

1from sklearn import metrics
2# predictions 
3pred_ovr = logR_ovr.predict(X)
4# Confusion matrix and classification report
5print("multi_class='ovr'\n")
6print(metrics.confusion_matrix(y,pred_ovr,labels=[0, 1, 2]))
7print(metrics.classification_report(y,pred_ovr,labels=[0, 1, 2]))

1from sklearn.linear_model import LogisticRegression
2# Creating model instances
3logR_mul = LogisticRegression(multi_class='multinomial') # multinomial
4# fitting both models 
5logR_mul.fit(X,y)
6# Accuracy score 
7print("Score when multi_class='multinomial':",logR_mul.score(X,y))

1from sklearn import metrics
2# predictions 
3pred_mul = logR_mul.predict(X)
4# Confusion matrix and classification report
5print("multi_class='multinomial'\n")
6print(metrics.confusion_matrix(y,pred_mul,labels=[0, 1, 2]))
7print(metrics.classification_report(y,pred_mul,labels=[0, 1, 2]))

1# probabilities when using one-vs-rest
2print(logR_ovr.predict_proba(X)[:2])
3# probabilities when using multinomial
4print(logR_mul.predict_proba(X)[:2])