Monitor.py 文件源码

def predict(self, nu, gamma):
        # classifier
        classifier = svm.OneClassSVM(kernel="rbf", nu=nu, gamma=gamma)
        # data for test
        from_timestamp = self.min_timestamp + datetime.timedelta(days=365)
        to_timestamp = self.max_timestamp
        test_timestamps, test_values = self.load_monitor_data(from_timestamp, to_timestamp, "nan")
        test_data = np.array(test_values)[:, 0:5]
        # data for train
        to_timestamp = self.min_timestamp + datetime.timedelta(days=365) + datetime.timedelta(hours=1)
        train_timestamps, train_values = self.load_monitor_data(self.min_timestamp, to_timestamp, "1")
        for i in range(len(test_timestamps)):
            # predict
            train_data = np.array(train_values)[:, 0:5]
            classifier.fit(train_data)
            label = classifier.predict(test_data[i])[0]
            test_values[i][5] = int(label)
            if label == 1:
                test_values[i][6] = 0.0
                train_values.append(test_values[i])
            else:
                test_values[i][6] = 1.0
            print test_timestamps[i], label, test_values[i]
        # write result into monitor file
        fr = open(self.monitor_file, "r")
        header = fr.readline()
        lines = fr.readlines()
        fr.close()
        fw = open(self.monitor_file, "w")  # update monitor file
        fw.write(header)
        for line in lines:
            timestamp = datetime.datetime.strptime(line.strip().split(",")[0], "%Y-%m-%d %H:%M:%S")
            if timestamp in test_timestamps:
                idx = test_timestamps.index(timestamp)
                value = test_values[idx]
                timestamp = str(timestamp)
                temperature = str(value[0])
                ph = str(value[1])
                conductivity = str(value[2])
                orp = str(value[3])
                do = str(value[4])
                label = str(int(value[5]))
                outlier_prob = str(value[6])
                event_prob = str(value[7])
                m = [timestamp, temperature, ph, conductivity, orp, do, label, outlier_prob, event_prob]
                fw.write(",".join(m) + "\n")
            else:
                fw.write(line)
        fw.close()