There is a dataset of real service calls of MT. The first level directory represents the day of march and the second represents the different interval(5 minutes for an interval) of each day.

Dataset1 only contains total service calls of three days (i.e., 10,19,22 March) and some real anomalies used in the last part "Example" exist in this dataset. Link : https://www.jianguoyun.com/p/De4m7c8QtbXZCBiktbID

• This dataset contains 31 files of Jan and each file contains all service calls of each day.
• The first number of each line of a file represents the interval. Link : https://www.jianguoyun.com/p/DTPfNicQtbXZCBiTtbID

Attention

• The dataset has been desensitized.
• ''A4,B0,C1,D1,E30,F9,G6055,2,200'' is a line of a file, and the first 7 values are the different dimensional values and the 8th value presents the number of service calls, further, the last is the status code of these service calls (200, the code for a successful service call, otherwise, failed).

The python file 'ImpAPTr.py' is the main body of our tool and you should run the file 'ImpAPTr_test.py'.

1. When you notice the DSR(Declining Success Rate) of SRSC(Success Rate of Service Calls), you should get the interval on where the DSR occurs.
2. Please run the file 'ImpAPTr_test.py' by the following command,

python ImpAPTr_test.py [day] [interval]

The parameter 'day' and 'interval' are the time of DSR's occuring.

3. After the running of the tool, there are some candidate clues which can benefit operators to find out the real 'root cause' and maintain the stability of service.

• /ImpAPTr_module/dataset/..
• /ImpAPTr_module/ImpAPTr.py
• /ImpAPTr_module/ImpAPTr_test.py

We propose two anomaly examples for the service on March. The first is an example of sharp DSR, and another is slight drop.

• 2020.3.10 08:00~08:05 --python ImpAPTr_test.py 10 480
• 2020.3.19 11:20~11:25 --python ImpAPTr_test.py 19 680