Training a logistic regression model with River

train_online_model.py

from river import linear_model, compose, preprocessing, metrics

# create logistic regression model with a standard scaler
model = compose.Pipeline(
    preprocessing.StandardScaler(),
    linear_model.LogisticRegression()
)

# use ROCAUC to evaluate our model
metric = metrics.ROCAUC()

# import the toy dataset for classifying phishing websites
dataset = datasets.Phishing() 

for x, y in dataset:
    y_pred = model.predict_proba_one(x) # get predicted probability
    model.learn_one(x, y) # train on a single example
    metric.update(y, y_pred) # update rocauc

print(metric)