fraud-detection-federated-learning/client.py at main · smridhi06/fraud-detection-federated-learning · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
from model import create_model
from utils import add_noise


class Client:
    def __init__(self, data, client_name="Client"):
        self.client_name = client_name
        self.X_train, self.X_test, self.y_train, self.y_test = data
        self.model = create_model()

    def train(self, global_weights=None):
        self.model.fit(self.X_train, self.y_train)

        if global_weights is not None:
            self.model.coef_ = global_weights[0].copy()
            self.model.intercept_ = global_weights[1].copy()

        self.model.fit(self.X_train, self.y_train)

        weights = (self.model.coef_.copy(), self.model.intercept_.copy())
        return add_noise(weights)

    def set_weights(self, weights):
        self.model.fit(self.X_train, self.y_train)
        self.model.coef_ = weights[0].copy()
        self.model.intercept_ = weights[1].copy()

    def evaluate(self):
        return self.model.score(self.X_test, self.y_test)

    def predict_proba(self, X):
        return self.model.predict_proba(X)

    def predict(self, X):
        return self.model.predict(X)