问基于PySpark的线性支持向量机多类分类

EN

import numpy as np
np.random.seed(323)

classes = [0, 1, 2, 3, 4, 5]

def make_point(classes):
    label = np.random.choice(classes)
    features = np.random.random(len(classes))
    features[label] += 10
    return LabeledPoint(label, features)

data  = sc.parallelize([make_point(classes) for _ in range(1000)])

xs = data.take(5)

def model_for_class(c, rdd):
    def adjust_label(lp):
        return LabeledPoint(1 if lp.label == c else 0, lp.features)

    model = SVMWithSGD.train(rdd.map(adjust_label))
    model.clearThreshold()
    return model

models = [model_for_class(c, data) for c in classes]

[(x.label, np.argmax([model.predict(x.features) for model in models]))
for x in xs]
## [(0.0, 0), (1.0, 1), (0.0, 0), (5.0, 5), (2.0, 2)]