AutoKAD/check.py at main · NetManAIOps/AutoKAD · 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
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
import os
import torch
import numpy as np
import random
import json
import pandas as pd

from detector.HW import train_and_predict as hw_train_and_predict
from detector.LSTM import train_and_predict as lstm_train_and_predict
from detector.VAE import train_and_predict as donut_train_and_predict

from utils.preprocess import minmax_scale

import matplotlib.pyplot as plt

def setup_seed(seed):
    torch.manual_seed(seed)
    torch.cuda.manual_seed_all(seed)
    np.random.seed(seed)
    random.seed(seed)
    torch.backends.cudnn.deterministic = True


def train_and_predict(train_df, test_df, params):
    algo_name = params['name']

    train_and_predict = None
    if algo_name == 'holt_winter':
        train_and_predict = hw_train_and_predict
    elif algo_name == 'LSTM':
        train_and_predict = lstm_train_and_predict
    elif algo_name == 'Donut':
        train_and_predict = donut_train_and_predict

    try:
        predicted = train_and_predict(train_df['value'], test_df['value'], params['params'])
    except Exception as e:
        print("params conflict!")
        print(e)

    # Make sure the prediction is one-dimension array
    predicted = predicted.reshape(-1)
    # replace nan with 0
    predicted[np.isnan(predicted)] = .0

    return predicted


def load_data(kpi):
    train_df = pd.read_csv(f"data/tzs/train/{kpi}.csv")
    test_df = pd.read_csv(f"data/tzs/test/{kpi}.csv")

    train_df['value'] = minmax_scale(train_df['value'])
    test_df['value'] = minmax_scale(test_df['value'])

    return train_df, test_df


def plot(ground_truth: np.ndarray, predicted: np.ndarray, save_path=None, labels=None, anomaly_score=None, figsize=(20, 5)):
    plot_num = 2
    if anomaly_score is not None:
        plot_num += 1

    plt.figure(figsize=(figsize[0], figsize[1] * plot_num))

    x = np.arange(0, len(ground_truth))

    # Test fig
    plt.subplot(plot_num, 1, 1)
    y = ground_truth
    plt.plot(x, y)
    plt.title("Test")

    if labels is not None:
        anomaly_pos = labels != 0
        plt.plot(x[anomaly_pos], y[anomaly_pos], marker='o', linestyle='None', color='r', markersize=2)

    # predicted fig
    plt.subplot(plot_num, 1, 2)
    y = predicted
    plt.plot(x, y)
    plt.title("Prediction")


    # Anomaly score
    plt.subplot(plot_num, 1, 3)
    y = anomaly_score
    plt.plot(x, y)
    plt.title("Anomaly Score")

    if save_path is None:
        plt.show()
    else:
        plt.savefig(save_path)


if __name__ == '__main__':
    setup_seed(329)

    kpi = "a8c06b47-cc41-3738-9110-12df0ee4c721"
    os.makedirs(f"plot/{kpi}",  exist_ok=True)

    check_file = "trials/msenf_tpe_100.json"
    trials = json.load(open(check_file, 'r'))

    train_df, test_df = load_data(kpi)

    cnt = 1

    for trial in trials:
        print(json.dumps(trial, indent=4, sort_keys=True))

        algo_name = trial['name']
        predicted = train_and_predict(train_df, test_df, trial)
        anomaly_score = abs(predicted - test_df['value'].to_numpy())


        # plot(ground_truth=test_df['value'].to_numpy(),
        #      predicted=predicted,
        #      labels=test_df['label'].to_numpy(),
        #      anomaly_score=anomaly_score,
        #      save_path=f"plot/{cnt}.png")

        cnt += 1