freeEDU/Log-Decoder

1"""2The implementation of the SVM model for anomaly detection.3 4Authors: 5    LogPAI Team6 7Reference: 8    [1] Yinglung Liang, Yanyong Zhang, Hui Xiong, Ramendra Sahoo. Failure Prediction 9        in IBM BlueGene/L Event Logs. IEEE International Conference on Data Mining10        (ICDM), 2007.11 12"""13 14import numpy as np15from sklearn import svm16from ..utils import metrics17 18class SVM(object):19 20    def __init__(self, penalty='l1', tol=0.1, C=1, dual=False, class_weight=None, 21                 max_iter=100):22        """ The Invariants Mining model for anomaly detection23        Arguments24        ---------25        See SVM API: https://scikit-learn.org/stable/modules/generated/sklearn.svm.LinearSVC.html26        27        Attributes28        ----------29            classifier: object, the classifier for anomaly detection30 31        """32        self.classifier = svm.LinearSVC(penalty=penalty, tol=tol, C=C, dual=dual, 33                                        class_weight=class_weight, max_iter=max_iter)34 35    def fit(self, X, y):36        """37        Arguments38        ---------39            X: ndarray, the event count matrix of shape num_instances-by-num_events40        """41        print('====== Model summary ======')42        self.classifier.fit(X, y)43 44    def predict(self, X):45        """ Predict anomalies with mined invariants46 47        Arguments48        ---------49            X: the input event count matrix50 51        Returns52        -------53            y_pred: ndarray, the predicted label vector of shape (num_instances,)54        """55        56        y_pred = self.classifier.predict(X)57        return y_pred58 59    def evaluate(self, X, y_true):60        print('====== Evaluation summary ======')61        y_pred = self.predict(X)62        precision, recall, f1 = metrics(y_pred, y_true)63        print('Precision: {:.3f}, recall: {:.3f}, F1-measure: {:.3f}\n'.format(precision, recall, f1))64        return precision, recall, f165