pyculiarity 时间序列（异常流量）异常检测初探——感觉还可以，和Facebook的fbprophet本质上一样

demo:

from pyculiarity import detect_ts
import matplotlib.pyplot as plt
import pandas as pd
import matplotlib
matplotlib.style.use('ggplot')

__author__ = 'willmcginnis'

if __name__ == '__main__':
    # first run the models
    twitter_example_data = pd.read_csv('./raw_data.csv', usecols=['timestamp', 'count'])
    results = detect_ts(twitter_example_data, max_anoms=0.05, alpha=0.001, direction='both', only_last=None)

    # format the twitter data nicely
    twitter_example_data['timestamp'] = pd.to_datetime(twitter_example_data['timestamp'])
    twitter_example_data.set_index('timestamp', drop=True)

    # make a nice plot
    f, ax = plt.subplots(2, 1, sharex=True)
    ax[0].plot(twitter_example_data['timestamp'], twitter_example_data['value'], 'b')
    ax[0].plot(results['anoms'].index, results['anoms']['anoms'], 'ro')
    ax[0].set_title('Detected Anomalies')
    ax[1].set_xlabel('Time Stamp')
    ax[0].set_ylabel('Count')
    ax[1].plot(results['anoms'].index, results['anoms']['anoms'], 'b')
    ax[1].set_ylabel('Anomaly Magnitude')
    plt.show()

demo2代码如下：

from matplotlib import pyplot as plt
from pyculiarity import detect_ts
import pandas as pd
import numpy as np

twitter_example_data = pd.read_csv('raw_data.csv',
                                    usecols=['timestamp', 'count'])

plt.plot(range(0, len(twitter_example_data)), twitter_example_data['count'],  "k.", label='points')
results = detect_ts(twitter_example_data,
                    max_anoms=0.02,
                    direction='both')
print(results['anoms'][0:10])
print(results['anoms'][-10:])
print(len(results['anoms']))

for timestamp, anomal_val in zip(results['anoms']['timestamp'], results['anoms']['anoms']):
    print(timestamp, anomal_val)
    index_list = np.where(twitter_example_data["timestamp"] == timestamp)
    assert len(index_list) == 1
    plt.plot([index_list[0]], anomal_val, "rX", label='abnormal points')
plt.show()

效果图：

原始数据图：

红色为检测出来的异常点：