CN115102835A

CN115102835A - ONU client side power-off intelligent judgment method based on machine learning

Info

Publication number: CN115102835A
Application number: CN202210493123.8A
Authority: CN
Inventors: 徐巍; 唐慧
Original assignee: Beijing Zznode Technology Co ltd
Current assignee: Beijing Zznode Technology Co ltd
Priority date: 2022-05-07
Filing date: 2022-05-07
Publication date: 2022-09-23

Abstract

An ONU client side power failure intelligent judgment method based on machine learning is characterized in that according to the phenomenon that ONU reports an alarm caused by the fact that part of ONU fixed time period customers automatically power off, a K-means clustering algorithm is utilized to analyze and obtain a model capable of efficiently and quickly detecting the ONU client side power off, the link that maintenance personnel manually dial a call to contact with customers to confirm whether the client automatically power off causes a fault alarm is omitted, whether the ONU power off alarm is suspected to be caused by the client side power off is judged in advance, and the ONU power off alarm is directly marked to be classified, so that the labor cost is reduced, the fault processing waiting time is shortened, the ONU power off fault workload is reduced, and the fault work order distribution accuracy and the fault processing efficiency are effectively improved.

Description

ONU client side power-off intelligent judgment method based on machine learning

Technical Field

The invention relates to the technical field of communication transmission services, in particular to an ONU client side power failure intelligent judgment method based on machine learning.

Background

At present, a large number of home broadband customers of a communication operator cut off power of a home Optical modem ONU (ONU, Optical Network Unit, Optical line terminal) at a fixed time, so that the ONU reports an alarm and triggers an operator fault center to dispatch a fault work order. In the fault, at present, a maintainer can only contact with a client by manually dialing a phone to confirm whether the fault is caused by power failure of the client, if the fault is caused by power failure of the client, the maintainer cannot process the fault, only the client goes home to reopen the switch, the fault can be automatically recovered, and the operation and maintenance staff can end the fault work order after the fault is recovered. The condition wastes the time for inquiring and returning the maintenance personnel faults and often causes the maintenance personnel to be examined because the maintenance personnel do not return the order in time.

In order to effectively solve the problems of long time consumption for processing a power failure work order of a domestic optical modem ONU and shortage of human resources, the invention provides an ONU client side power failure intelligent detection method based on a K-means clustering algorithm (K-means is also called as K mean value, and is a clustering algorithm) of machine learning by taking machine learning as a basic background, and aims to solve the problem that the conventional ONU power failure cannot be judged to be the client side or the operator side by utilizing the characteristics of high clustering efficiency and good clustering effect when the machine learning is used for carrying out clustering analysis on a large-scale data set, and realize the intelligent judgment of the ONU client side power failure.

Disclosure of Invention

Aiming at the defects in the prior art, the invention provides an ONU client side power failure intelligent judgment method based on machine learning, according to the phenomenon that ONU reports an alarm caused by that part of ONU fixed time period customers automatically power off, a K-means clustering algorithm is utilized to analyze and obtain a model capable of efficiently and quickly detecting the ONU client side power off, the link that maintenance personnel manually dial a call to contact with the customers to confirm whether the customer automatically power off causes a fault alarm is omitted, whether the ONU power off alarm is suspected to be caused by the customer side power off is judged in advance, and the ONU power off fault work order is directly marked to be classified, so that the labor cost is reduced, the fault processing waiting time is reduced, the ONU power off fault work order quantity is reduced, and the fault work order distribution accuracy and the processing efficiency are effectively improved.

The technical solution of the invention is as follows:

an ONU client side power failure intelligent judgment method based on machine learning is characterized in that according to the phenomenon that ONU reports an alarm caused by the fact that a client self-power failure exists in a fixed time interval in ONU operation and maintenance, a K-means clustering algorithm is utilized, a model for detecting the ONU client side power failure is obtained through analysis, whether the ONU power failure alarm is suspected to be caused by the client side power failure is judged in advance, and direct marking is carried out for classification processing, so that the link that maintenance personnel manually dial a call to contact with the client to confirm whether the client self-power failure alarm is caused can be omitted, the labor cost is reduced, the fault processing waiting time is shortened, the ONU power failure fault work order quantity is reduced, and the fault work order distribution accuracy and the processing efficiency are effectively improved.

The method comprises the following steps:

step 1, collecting ONU power-down alarm;

step 2, carrying out resource positioning aiming at ONU power failure alarm;

step 3, identifying resources and alarm time periods through a K-means machine learning model;

step 4, judging whether the resources are consistent and meet the suspected client side outage time period, if so, entering step 8 after sequentially going through steps 5a to 7a, and if not, entering step 8 after sequentially going through steps 5b to 7 b;

step 5a, marking a suspected client side power-off mark;

step 6a, dispatching a notification work order;

step 7a, the maintenance personnel receive orders and fill in the predicted recovery time;

step 5b, undoubtedly, a client side power-off mark is similar;

step 6b, dispatching a fault work order;

step 7b, the maintenance personnel receive orders and process faults, and the clearing alarm is received;

step 8, returning the order;

and 9, archiving.

The ONU power-down alarm in the step 1 comprises one or more of the following: the branch optical fiber is broken or the OLT cannot detect the expected optical signal LOSi/LOBi of the ONT; ONU in GPON alarm is disconnected; ONU signals in GPON alarm are invalid; ONU signal loss in GPON alarm; LINK _ LOSS in LINK alarm.

And the step 2 comprises that the operator fault center acquires resource data through an interface with a resource management system, whether the equipment attributive customer information and the area information matched with the ONU power failure alarm are complete or not is positioned, if the equipment attributive customer information and the area information are incomplete, the process is ended, and if the equipment attributive customer information and the area information are complete, the step 3 is executed.

And the step 3 comprises the steps of transmitting the ONU name, the attribution customer information, the attribution area information and the alarm occurrence time of the ONU power failure alarm into a K-means machine learning model, detecting and identifying alarm data by the model, alarming when the resource is consistent and meets the suspected client side power failure time period, and marking a suspected client side power failure mark.

The step 3 comprises the following steps:

step K1, collecting ONU power-down history alarm;

k2, performing sampling analysis on ONU power-down history alarms;

k3, selecting a K-means clustering algorithm according to a time sequence according to sampling analysis;

k4, calculating by a K-means clustering algorithm to obtain a clustering cluster;

and K5, storing the ONU network element resources and the client side power-off period information according to the calculation result.

The step K1 includes extracting the resource, occurrence time and recovery time of the ONU power failure alarm from the fault center historical alarm library, extracting the fault reason of the fault work order corresponding to the alarm, and analyzing that the number of ONU samples with the order distribution quantity exceeding 50 exceeds 160, or the number of ONU samples with the order distribution quantity exceeding 20 exceeds 1000.

The step K2 includes performing scatter diagram processing on the occurrence time and recovery time of the corresponding alarm of which the failure cause is the power failure of the client side, so as to find out the regularity of the occurrence time and recovery time of the power failure alarm of part of ONUs.

The K-means clustering algorithm in the step K3 adopts the following algorithm formula:

k represents k clustering centers, the clustering centers are set according to time periods, ci represents the ith clustering center, i is a positive integer, dist represents the Euclidean distance, X represents objects belonging to ci, X represents the number of the objects in ci, and min is a minimum value algorithm.

In the step K4, each cluster has a center point parameter and a cluster radius parameter.

The step K5 includes analyzing the ONU with the highest dispatching frequency every day through the continuously accumulated training sample data, excavating and updating the time period of the suspected user power failure habit, comparing the real-time alarm with the time period, dispatching the alarm meeting the time period to notify the work order, and dispatching the alarm not meeting the time period to send the fault work order.

The invention has the following technical effects: the invention relates to an ONU client side power failure intelligent judgment method based on machine learning, which aims at the problem that the ONU power failure at the present stage is determined whether the fault is caused by the power failure of a client or not by dispatching a fault work order and contacting the client by manually dialing a telephone after a maintenance worker receives the order, the method has the advantages that the invested labor cost is high, the consumed time is long, and the like, according to the characteristic that the ONU reports the alarm caused by the self power-off of part of the ONU at the fixed time interval, a K-means clustering algorithm is used for analyzing and obtaining a model capable of efficiently and quickly detecting the power-off of the ONU at the client side, the link of manually calling and connecting with the client is saved, by the detection model of the method, whether ONU power-off alarm is suspected to be caused by the power-off of the client side is directly marked, so that the labor cost is reduced, the fault processing waiting time is reduced, the ONU power-off fault work order quantity is reduced, and the fault work order distribution accuracy and the fault work order processing efficiency are effectively improved.

The invention has the following characteristics: 1. the invention introduces a machine learning clustering algorithm as a method for replacing an artificial telephone and confirming whether the power is off by a client. 2. And a machine learning technology selection mechanism of the ONU client side power-off intelligent judgment method based on machine learning. 3. And calculating to obtain the clustering method through a K-means algorithm.

Drawings

Fig. 1 is a schematic flowchart of an ONU client side power-off intelligent determination method based on machine learning according to the present invention. An ONU is an Optical Network Unit (Optical Network Unit). FIG. 1 includes step 1, where an ONU power down alarm is collected; step 2, whether the resource can be positioned or not is judged, if not, the process is ended, and if yes, the process enters the step 3; step 3, the K-means machine learning model identifies resources and alarm time periods (the K-means is also called K mean value and is a clustering algorithm); step 4, judging whether the resources are consistent and meet the suspected client side outage time interval, if so, sequentially going through the step 5a to the step 7a and then going to the step 8, and if not, sequentially going through the step 5b to the step 7b and then going to the step 8; step 5a, marking a suspected client side power-off mark; step 6a, dispatching a notification work order; step 7a, the maintenance personnel receive orders and fill in the predicted recovery time; step 5b, undoubtedly, a client side power-off mark is similar; step 6b, dispatching a fault work order; step 7b, the maintenance personnel receive orders and process faults, and wait for receiving a clearing alarm; step 8, returning the order; and 9, archiving.

FIG. 2 is a schematic diagram of a resource and alarm period identification process based on a K-means machine learning model. Fig. 2 includes a step K1, collecting ONU power-down history alarm; k2, performing ONU power failure history alarm sampling analysis; k3, selecting a K-means clustering algorithm according to a time sequence according to sampling analysis; k4, calculating by a K-means algorithm to obtain a cluster; and K5, storing the ONU network element resources and the client side power-off period information according to the calculation result.

Fig. 3 is a schematic diagram of ONU power down alarm period dispersion of a first sample ONU1-FTTH of ONU power down history alarm sampling analysis. In fig. 3, The abscissa is date (year/month/day), The ordinate is time (hour/minute/second), dark dots (original is blue dots) indicate occurrence time, light dots (original is red dots) indicate clearing time, and FTTH is Fiber To The Home (Fiber To The Home).

Fig. 4 is a schematic diagram of ONU power down alarm period dispersion of a second sample ONU power down history alarm sampling analysis ONU 2-FTTH. In fig. 4, The abscissa is date (year/month/day), The ordinate is time (hour/minute/second), dark dots (original is blue dots) indicate occurrence time, light dots (original is red dots) indicate clearing time, and FTTH is Fiber To The Home (Fiber To The Home).

Fig. 5 is a schematic diagram of ONU power down alarm period dispersion of a third sample ONU3-FTTH of ONU power down history alarm sampling analysis. In fig. 5, The abscissa is The date (year/month/day), The ordinate is The time (hour/minute/second), The dark color point (The original is a blue color point) indicates The occurrence time, The light color point (The original is a red color point) indicates The clearing time, and FTTH is The Fiber To The Home (Fiber To The Home).

Fig. 6 is a schematic diagram of ONU power down alarm period dispersion points of a fourth sample ONU4-FTTH of ONU power down history alarm sampling analysis. In fig. 6, The abscissa is The date (year/month/day), The ordinate is The time (hour/minute/second), The dark color point (The original is a blue color point) indicates The occurrence time, The light color point (The original is a red color point) indicates The clearing time, and FTTH is The Fiber To The Home (Fiber To The Home).

Fig. 7 is a schematic diagram of ONU power down alarm period dispersion of a fifth sample ONU power down history alarm sampling analysis ONU 5-FTTH. In fig. 7, The abscissa is The date (year/month/day), The ordinate is The time (hour/minute/second), The dark color point (The original is a blue color point) indicates The occurrence time, The light color point (The original is a red color point) indicates The clearing time, and FTTH is The Fiber To The Home (Fiber To The Home).

Detailed Description

The invention is explained below with reference to the figures (fig. 1-7).

Fig. 1 is a schematic flowchart of an ONU client side power-off intelligent determination method based on machine learning according to the present invention. FIG. 2 is a schematic diagram of a resource and alarm period identification process based on a K-means machine learning model. Referring to fig. 1 to 2, an ONU client side power failure intelligent judgment method based on machine learning analyzes a model for detecting ONU client side power failure to judge in advance whether ONU power failure alarm is suspected to be caused by client side power failure according to a phenomenon that the ONU reports alarm due to client self power failure at a fixed time interval in operation and maintenance, and directly performs marking for classification, so that a link of manually dialing a call by a maintainer to contact with a client to confirm whether the client self power failure causes fault alarm can be omitted, labor cost is reduced, fault processing waiting time is reduced, ONU power failure fault workload is reduced, and fault work order distribution accuracy and processing efficiency are effectively improved.

The method comprises the following steps: step 1, collecting ONU power-down alarm; step 2, carrying out resource positioning aiming at ONU power failure alarm; step 3, identifying resources and alarm time periods through a K-means machine learning model; step 4, judging whether the resources are consistent and meet the suspected client side outage time period, if so, entering step 8 after sequentially going through steps 5a to 7a, and if not, entering step 8 after sequentially going through steps 5b to 7 b; step 5a, marking a suspected client side power-off mark; step 6a, dispatching a notification work order; step 7a, a maintenance person receives an order and fills in the expected recovery time; step 5b, undoubtedly, a client side power-off mark is similar; step 6b, dispatching a fault work order; step 7b, the maintenance personnel receive orders and process faults, and the clearing alarm is received; step 8, returning the order; and 9, archiving.

The ONU power-down alarm in the step 1 comprises one or more of the following: the branch optical fiber is broken or the OLT cannot detect the expected optical signal LOSi/LOBi of the ONT; ONU in GPON alarm is disconnected; ONU signals in GPON alarm are invalid; ONU signal loss in GPON alarm; LINK _ LOSS in LINK alarm. And the step 2 comprises that the operator fault center acquires resource data through an interface with a resource management system, whether the equipment attributive customer information and the area information matched with the ONU power failure alarm are complete or not is positioned, if the equipment attributive customer information and the area information are incomplete, the process is ended, and if the equipment attributive customer information and the area information are complete, the step 3 is executed. And the step 3 comprises the steps of transmitting the ONU name, the attribution customer information, the attribution area information and the alarm occurrence time of the ONU power failure alarm into a K-means machine learning model, detecting and identifying alarm data by the model, alarming when the resource is consistent and meets the suspected customer side power failure time period, and marking a suspected customer side power failure mark.

The step 3 comprises the following steps: step K1, collecting ONU power failure history alarm; k2, performing sampling analysis on ONU power-down history alarms; k3, selecting a K-means clustering algorithm according to a time sequence according to sampling analysis; k4, calculating by a K-means clustering algorithm to obtain a clustering cluster; and K5, storing the ONU network element resources and the client side power-off period information according to the calculation result. The step K1 includes extracting the resource, occurrence time and recovery time of the ONU power failure alarm from the fault center historical alarm library, extracting the fault reason of the fault work order corresponding to the alarm, and analyzing that the number of ONU samples with the order distribution quantity exceeding 50 exceeds 160, or the number of ONU samples with the order distribution quantity exceeding 20 exceeds 1000. The step K2 includes performing scatter diagram processing on the occurrence time and recovery time of the corresponding alarm of which the failure cause is the power failure of the client side, so as to find out the regularity of the occurrence time and recovery time of the power failure alarm of part of ONUs.

wherein k represents k clustering centers, the clustering centers are set according to time periods, ci represents the ith clustering center, i is a positive integer, dist represents the Euclidean distance, X represents objects belonging to ci, X represents the number of the objects in ci, and min is a minimum value-taking algorithm.

In the step K4, each cluster has a center point parameter and a cluster radius parameter. The step K5 includes analyzing the ONU with the highest dispatching frequency every day through the continuously accumulated training sample data, excavating and updating the time period of the suspected user power failure habit, comparing the real-time alarm with the time period, dispatching the alarm meeting the time period to notify the work order, and dispatching the alarm not meeting the time period to send the fault work order.

The invention provides an ONU client side power-off intelligent judgment method based on machine learning, which comprises the following specific detection steps (as shown in figure 1):

1. and (3) collecting ONU power failure alarm: the operator fault center collects alarms (PON, Passive Optical Network, GPON, gigabit Passive Optical Network) through a Network management interface of PON Network equipment, and enters an ONU client side power-off intelligent judgment flow when the ONU power-off alarms are collected. The ONU power failure alarm comprises the following alarm names: "the branch optical fiber is broken or the OLT cannot detect the expected optical signal (LOSi/LOBi) of the ONT", "GPON alarm ] ONU is disconnected", "[ GPON alarm ] ONU signal is failed", "[ GPON alarm ] ONU signal is lost", and "LINK _ LOSS".

2. Resource positioning: and the operator fault center acquires the resource data through the interface with the resource management system, and positions whether the equipment attributive customer information and the area information matched with the ONU power failure alarm are complete or not. If not, the flow is ended, and if not, the next step is continued.

3. The K-means machine learning model identifies resources and alarm time periods: and transmitting the ONU name, the attribution customer information, the attribution area information and the alarm occurrence time of the ONU power failure alarm into a K-means machine learning model, detecting and identifying alarm data by the model, and marking a suspected client side power failure mark for the alarm with consistent resources and meeting the suspected client side power failure time period.

4. Dispatching a notification work order: and sending a notification work order aiming at the ONU power failure alarm marked with the suspected client side power failure mark, and directly filling the expected recovery time and then returning the order and filing after the maintenance personnel receives the order.

5. Dispatching a fault work order: and dispatching a fault work order aiming at ONU power failure alarm without a suspected client side power failure mark, after a maintenance person receives the order, and after the maintenance person receives the order and processes the fault, waiting for receiving a clearing alarm, and then executing a receipt and filing.

The resource and alarm time interval identification process based on the K-means machine learning model is as follows (as shown in FIG. 2):

1. collecting ONU power failure history alarm: and extracting the resource, the occurrence time and the recovery time of the ONU power failure alarm from the fault center historical alarm library, and extracting the fault reason corresponding to the alarm related fault work order. And analyzing that the number of ONUs with the dispatch amount exceeding 50 exceeds 160, and the number of ONUs with the dispatch amount exceeding 20 exceeds 1000.

2. ONU power failure history alarm sampling analysis: and (3) processing the occurrence time and the recovery time of the corresponding alarm of the fault reason caused by the power failure of the client side through a scatter diagram, and finding that strong regularity exists in the occurrence time and the recovery time of part of the ONUs.

Sampling and analyzing data (more than 50 same ONU orders), wherein the ONU sample analysis comprises the following steps: ONU1-FTTH (Fiber To The Home), ONU2-FTTH, ONU3-FTTH, ONU4-FTTH and ONU5-FTTH, and analyzing ONU power failure alarm period scatter diagrams are shown in figures 3 To 7.

3. Selecting a K-means clustering algorithm according to a time sequence according to sampling analysis:

according to the characteristic that the client side is periodically powered off, selecting a K-means clustering algorithm according to a time sequence as a machine

The basic algorithm of learning is as follows:

k denotes k cluster centers, ci denotes the first center, and dist denotes the Euclidean distance.

4. Calculating by a K-means algorithm to obtain a cluster:

based on the sample analysis, it is found that the customer outage can be roughly divided into a plurality of time periods of 0-10 points, 10-14 points, 14-18 points, 18 points-20 points, 20-24 points, etc., so it is preliminarily suggested that the k value of the cluster center is set to 5. The formula input parameter is the occurrence time point of power failure alarm of each sample ONU, the data is obtained from a fault center historical alarm library, and the more samples are, the higher the analysis reliability is. According to the algorithm: center point and cluster radius of 5 cluster clusters 1.

5. And storing ONU network element resources and client side outage period information according to the calculation result:

and analyzing the ONU with the highest dispatching single frequency every day through the continuously accumulated training sample data, excavating and updating a time period of suspected user power failure habits, comparing the real-time alarm with the time period, and dispatching the alarm meeting the time period to inform the work order and the alarm not meeting the time period to dispatch the fault work order.

The concrete process of identifying resources and alarm time intervals based on the K-means machine learning model is described by combining with an actual case as follows:

taking ONU1-FTTH, ONU power down alarm (alarm name: branch optical fiber is broken or OLT can not detect the anticipated optical signal (LOSi/LOBi) of ONT) from 1 month and 1 day in 2021 to 6 months and 16 days in 2021 as samples, totally selecting 322 samples, wherein the sample data is as follows:

2021-06-16 12:15:14；2021-06-15 20:15:32；2021-06-15 12:30:26；2021-06-14 19:14:34；2021-06-14 18:55:13；2021-06-14 12:56:19；2021-06-13 19:03:46；2021-06-13 12:12:02；2021-06-12 19:57:41；2021-06-12 11:33:06；2021-06-11 20:08:19；2021-06-11 12:10:35；2021-06-10 20:30:00；2021-06-10 12:03:11；2021-06-09 20:24:22；2021-06-09 11:33:06；2021-06-09 02:46:54；2021-06-07 21:00:05；2021-06-07 15:56:47；2021-06-06 20:38:30；2021-06-06 12:10:25；2021-06-05 19:42:58；2021-06-05 12:40:38；2021-06-04 20:23:41；2021-06-03 20:29:29；2021-06-03 12:04:04；2021-06-02 20:32:07；2021-06-02 12:38:34；2021-06-01 20:44:28；2021-06-01 12:12:24；2021-05-31 20:17:42；2021-05-31 12:28:10；2021-05-30 21:22:15；2021-05-30 15:17:37；2021-05-30 12:00:51；2021-05-29 20:10:17；2021-05-29 11:54:17；2021-05-28 20:03:49；2021-05-28 11:50:11；2021-05-27 19:43:12；2021-05-27 12:35:13；2021-05-26 20:40:55；2021-05-26 12:00:52；2021-05-26 11:06:34；2021-05-25 20:34:53；2021-05-25 12:30:51；2021-05-24 20:29:35；2021-05-24 12:11:29；2021-05-23 20:37:44；2021-05-23 12:27:30；2021-05-22 20:22:45；2021-05-22 12:05:32；2021-05-21 20:32:40；2021-05-21 12:42:05；2021-05-20 20:14:21；2021-05-20 12:03:45；2021-05-19 20:30:52；2021-05-19 12:32:12；2021-05-18 20:30:06；2021-05-18 12:50:08；2021-05-17 20:19:10；2021-05-17 12:43:54；2021-05-16 20:39:40；2021-05-15 20:46:36；2021-05-15 12:40:55；2021-05-15 03:51:08；2021-05-13 20:28:33；2021-05-13 12:31:31；2021-05-13 11:39:24；2021-05-12 20:53:51；2021-05-12 13:17:21；2021-05-11 21:28:43；2021-05-11 12:47:08；2021-05-10 20:27:49；2021-05-10 11:55:44；2021-05-09 20:39:57；2021-05-09 12:32:11；021-05-08 20:28:05；2021-05-07 21:14:03；2021-05-07 13:23:19；2021-05-06 20:27:20；2021-05-06 13:29:41；2021-05-05 21:06:18；2021-05-05 13:20:55；2021-05-04 20:25:38；2021-05-03 20:10:29；2021-05-03 12:28:19；2021-05-02 20:30:57；2021-05-02 12:29:15；2021-05-01 20:33:26；2021-04-30 20:31:10；2021-04-30 12:59:58；2021-04-29 20:30:48；2021-04-29 13:06:42；2021-04-28 20:11:22；2021-04-28 12:27:51；2021-04-27 21:08:51；2021-04-26 19:22:44；2021-04-26 12:12:03；2021-04-25 20:30:12；2021-04-25 12:57:21；2021-04-24 20:09:00；2021-04-24 12:19:40；2021-04-23 20:39:22；2021-04-23 12:49:31；2021-04-20 21:02:22；2021-04-20 12:49:29；2021-04-19 22:00:19；2021-04-19 12:31:43；2021-04-19 06:06:24；2021-04-18 12:14:12；2021-04-17 20:17:07；2021-04-17 13:06:14；2021-04-16 20:17:16；2021-04-16 13:02:22；2021-04-15 20:10:28；2021-04-15 12:30:56；2021-04-14 20:12:31；2021-04-14 12:34:48；2021-04-13 20:13:37；2021-04-13 12:46:34；2021-04-12 20:05:42；2021-04-12 12:56:03；2021-04-11 20:34:19；2021-04-11 12:48:44；2021-04-10 20:16:03；2021-04-10 12:29:43；2021-04-09 20:07:10；2021-04-09 12:29:01；2021-04-08 20:26:53；2021-04-08 12:25:36；2021-04-06 20:18:27；2021-04-06 13:00:36；2021-04-05 20:00:20；2021-04-05 12:31:50；2021-04-04 20:58:35；2021-04-03 20:33:00；2021-04-03 13:02:22；2021-04-03 11:36:03；2021-04-02 20:36:00；2021-04-02 12:22:24；2021-04-01 19:44:28；2021-04-01 12:59:46；2021-03-31 21:23:00；2021-03-31 13:35:05；2021-03-29 19:05:41；2021-03-29 18:34:27；2021-03-29 12:26:03；2021-03-28 20:32:05；2021-03-28 12:18:47；2021-03-27 19:58:46；2021-03-27 12:19:28；2021-03-26 20:34:45；2021-03-26 12:53:30；2021-03-25 19:45:13；2021-03-25 12:57:23；2021-03-24 20:16:15；2021-03-24 12:54:04；2021-03-24 11:40:34；2021-03-23 20:25:24；2021-03-23 13:09:21；2021-03-22 20:05:56；2021-03-22 13:01:07；2021-03-21 20:32:30；2021-03-21 19:46:48；2021-03-21 13:36:51；2021-03-20 20:01:45；2021-03-20 17:52:59；2021-03-20 12:57:04；2021-03-20 11:41:46；2021-03-20 01:55:08；2021-03-19 13:27:35；2021-03-18 20:03:51；2021-03-17 19:47:49；2021-03-17 14:29:13；2021-03-16 19:03:06；2021-03-16 12:30:33；2021-03-16 12:30:33；2021-03-15 19:51:17；2021-03-15 12:49:02；2021-03-14 20:32:13；2021-03-14 12:44:03；2021-03-13 20:24:32；2021-03-13 12:24:47；2021-03-12 20:13:00；2021-03-12 11:45:59；2021-03-11 20:00:34；2021-03-11 12:29:03；2021-03-10 20:05:31；2021-03-10 12:28:01；2021-03-09 20:29:57；2021-03-09 12:55:26；2021-03-08 20:57:35；2021-03-08 19:30:23；2021-03-08 12:21:38；2021-03-07 20:19:04；2021-03-07 12:42:15；2021-03-06 18:09:17；2021-03-06 12:26:03；2021-03-05 20:00:19；2021-03-05 12:39:41；2021-03-04 19:56:13；2021-03-04 12:12:31；2021-03-03 20:09:03；2021-03-03 12:36:46；2021-03-02 20:17:45；2021-03-02 12:33:29；2021-03-02 11:42:36；2021-03-01 20:45:11；2021-03-01 13:15:54；2021-02-28 20:34:23；2021-02-28 12:25:13；2021-02-27 20:42:39；2021-02-27 13:12:47；2021-02-26 20:00:20；2021-02-26 12:51:10；2021-02-25 19:57:12；2021-02-25 12:25:31；2021-02-24 19:56:45；2021-02-24 12:35:45；2021-02-23 20:11:53；2021-02-23 12:35:53；2021-02-22 19:55:23；2021-02-22 12:43:57；2021-02-21 20:35:32；2021-02-21 12:27:53；2021-02-21 11:28:33；2021-02-20 20:08:16；2021-02-19 20:18:15；2021-02-19 12:49:26；2021-02-18 19:44:59；2021-02-18 12:29:26；2021-02-17 19:58:11；2021-02-17 13:26:39；2021-02-16 20:09:23；2021-02-16 12:27:45；2021-02-15 20:31:05；2021-02-15 12:46:54；2021-02-14 20:59:14；2021-02-14 12:51:34；2021-02-13 19:55:08；2021-02-12 19:53:36；2021-02-12 12:28:13；2021-02-11 19:57:44；2021-02-11 12:56:47；2021-02-10 19:53:05；2021-02-09 18:43:57；2021-02-09 12:40:07；2021-02-08 18:51:35；2021-02-08 12:37:51；2021-02-07 20:25:03；2021-02-07 12:19:54；2021-02-06 20:01:41；2021-02-06 12:41:39；2021-02-05 20:11:41；2021-02-05 12:14:31；2021-02-04 19:28:05；2021-02-04 12:22:21；2021-02-03 19:44:01；2021-02-03 13:47:09；2021-02-02 20:01:23；2021-02-02 13:29:17；2021-02-01 19:38:30；2021-02-01 12:58:59；2021-01-31 20:35:47；2021-01-31 12:42:44；2021-01-30 19:17:52；2021-01-30 12:45:37；2021-01-29 20:18:04；2021-01-29 13:04:09；2021-01-28 20:08:25；2021-01-28 12:40:04；2021-01-28 07:27:54；2021-01-27 12:29:06；2021-01-26 18:53:26；2021-01-26 12:27:38；2021-01-25 20:23:58；2021-01-25 12:56:47；2021-01-24 20:31:53；2021-01-24 12:35:29；2021-01-23 19:38:34；2021-01-23 12:41:13；2021-01-22 20:20:46；2021-01-21 18:55:52；2021-01-21 12:31:26；2021-01-20 20:04:42；2021-01-19 19:57:57；2021-01-19 12:48:08；2021-01-18 19:56:37；2021-01-18 13:12:55；2021-01-17 20:00:38；2021-01-17 12:02:59；2021-01-16 19:37:13；2021-01-16 12:15:09；2021-01-15 20:09:23；2021-01-15 13:26:43；2021-01-14 20:14:17；2021-01-14 12:12:10；2021-01-13 19:57:16；2021-01-13 13:23:40；2021-01-12 19:52:08；2021-01-11 19:51:10；2021-01-11 12:26:15；2021-01-10 20:30:14；2021-01-10 12:32:45；2021-01-09 19:39:25；2021-01-09 12:56:02；2021-01-09 12:03:24；2021-01-08 19:32:03；2021-01-08 12:39:13；2021-01-08 07:15:14；2021-01-07 12:29:53；2021-01-06 20:06:01；2021-01-06 12:27:37；2021-01-05 19:59:11；2021-01-05 13:00:43；2021-01-04 20:10:06；2021-01-04 12:45:47；2021-01-03 20:05:24；2021-01-03 12:55:38；2021-01-02 19:28:09；2021-01-02 12:55:15；2021-01-01 19:56:08；2021-01-01 12:55:12。

5 clustering clusters are obtained by calculating through a K-means algorithm (the K value is set to be 5), and the clustering clusters are respectively as follows:

clustering 1: the center point was 20:15:44 and the cluster radius was 769.47015, i.e., 20:02:54-20:28:33, satisfying the example number 112.

Clustering 2: the center point was 04:53:47 and the cluster radius was 7363.6665, i.e., 02:51:03-06:56:40, satisfying the example number 6.

Clustering cluster 3: the center point was 19:15:03 and the cluster radius was 1443.6228, i.e., 18:51:00-19:39:07, satisfying the example number 28.

Clustering 4: the center point is 12:39:10 and the cluster radius is 1434.0354, i.e., 12:15:16-13:03:04, satisfying the example number 152.

Clustering cluster 5: the center point was 21:04:10 and the cluster radius was 895.44446, i.e., 20:49:14-21:19:05, satisfying the example number 18.

The detailed results are as follows:

Cluster

Cluster_id＝1,center:{Point_id＝-1[72944.17]clusterId:0 dist:0.0}

Point_id＝6[71861.0]clusterId:1 dist:1083.1719

Point_id＝9[72499.0]clusterId:1 dist:445.17188

Point_id＝10[73800.0]clusterId:1 dist:855.8281

Point_id＝14[73462.0]clusterId:1 dist:517.8281

Point_id＝17[74310.0]clusterId:1 dist:1365.8281

Point_id＝21[73421.0]clusterId:1 dist:476.82812

Point_id＝23[73769.0]clusterId:1 dist:824.8281

Point_id＝24[73927.0]clusterId:1 dist:982.8281

Point_id＝29[73062.0]clusterId:1 dist:117.828125

Point_id＝33[72617.0]clusterId:1 dist:327.17188

Point_id＝36[72229.0]clusterId:1 dist:715.1719

Point_id＝43[74093.0]clusterId:1 dist:1148.8281

Point_id＝45[73775.0]clusterId:1 dist:830.8281

Point_id＝47[74264.0]clusterId:1 dist:1319.8281

Point_id＝49[73365.0]clusterId:1 dist:420.82812

Point_id＝50[73960.0]clusterId:1 dist:1015.8281

Point_id＝52[72861.0]clusterId:1 dist:83.171875

Point_id＝55[73852.0]clusterId:1 dist:907.8281

Point_id＝56[73806.0]clusterId:1 dist:861.8281

Point_id＝59[73150.0]clusterId:1 dist:205.82812

Point_id＝60[74380.0]clusterId:1 dist:1435.8281

Point_id＝66[73713.0]clusterId:1 dist:768.8281

Point_id＝71[73669.0]clusterId:1 dist:724.8281

Point_id＝73[74397.0]clusterId:1 dist:1452.8281

Point_id＝75[73685.0]clusterId:1 dist:740.8281

Point_id＝78[73640.0]clusterId:1 dist:695.8281

Point_id＝82[73538.0]clusterId:1 dist:593.8281

Point_id＝84[72629.0]clusterId:1 dist:315.17188

Point_id＝85[73857.0]clusterId:1 dist:912.8281

Point_id＝87[74006.0]clusterId:1 dist:1061.8281

Point_id＝88[73870.0]clusterId:1 dist:925.8281

Point_id＝91[73848.0]clusterId:1 dist:903.8281

Point_id＝92[72682.0]clusterId:1 dist:262.17188

Point_id＝97[73812.0]clusterId:1 dist:867.8281

Point_id＝100[72540.0]clusterId:1 dist:404.17188

Point_id＝102[74362.0]clusterId:1 dist:1417.8281

Point_id＝110[73027.0]clusterId:1 dist:82.828125

Point_id＝112[73036.0]clusterId:1 dist:91.828125

Point_id＝113[72628.0]clusterId:1 dist:316.17188

Point_id＝115[72751.0]clusterId:1 dist:193.17188

Point_id＝118[72817.0]clusterId:1 dist:127.171875

Point_id＝119[72342.0]clusterId:1 dist:602.1719

Point_id＝121[74059.0]clusterId:1 dist:1114.8281

Point_id＝123[72963.0]clusterId:1 dist:18.828125

Point_id＝125[72430.0]clusterId:1 dist:514.1719

Point_id＝128[73613.0]clusterId:1 dist:668.8281

Point_id＝129[73107.0]clusterId:1 dist:162.82812

Point_id＝131[72020.0]clusterId:1 dist:924.1719

Point_id＝135[73980.0]clusterId:1 dist:1035.8281

Point_id＝138[74160.0]clusterId:1 dist:1215.8281

Point_id＝146[73925.0]clusterId:1 dist:980.8281

Point_id＝148[71926.0]clusterId:1 dist:1018.1719

Point_id＝151[74085.0]clusterId:1 dist:1140.8281

Point_id＝155[72975.0]clusterId:1 dist:30.828125

Point_id＝157[73524.0]clusterId:1 dist:579.8281

Point_id＝160[72356.0]clusterId:1 dist:588.1719

Point_id＝162[71208.0]clusterId:1 dist:1736.1719

Point_id＝163[73950.0]clusterId:1 dist:1005.8281

Point_id＝168[72105.0]clusterId:1 dist:839.1719

Point_id＝170[72231.0]clusterId:1 dist:713.1719

Point_id＝171[71269.0]clusterId:1 dist:1675.1719

Point_id＝176[71477.0]clusterId:1 dist:1467.1719

Point_id＝178[73933.0]clusterId:1 dist:988.8281

Point_id＝181[73472.0]clusterId:1 dist:527.8281

Point_id＝183[72780.0]clusterId:1 dist:164.17188

Point_id＝184[72034.0]clusterId:1 dist:910.1719

Point_id＝186[72331.0]clusterId:1 dist:613.1719

Point_id＝189[73797.0]clusterId:1 dist:852.8281

Point_id＝194[73144.0]clusterId:1 dist:199.82812

Point_id＝198[72019.0]clusterId:1 dist:925.1719

Point_id＝199[71773.0]clusterId:1 dist:1171.1719

Point_id＝201[72543.0]clusterId:1 dist:401.17188

Point_id＝205[73065.0]clusterId:1 dist:120.828125

Point_id＝208[74063.0]clusterId:1 dist:1118.8281

Point_id＝214[71832.0]clusterId:1 dist:1112.1719

Point_id＝217[71805.0]clusterId:1 dist:1139.1719

Point_id＝218[72713.0]clusterId:1 dist:231.17188

Point_id＝221[71723.0]clusterId:1 dist:1221.1719

Point_id＝223[74132.0]clusterId:1 dist:1187.8281

Point_id＝225[72496.0]clusterId:1 dist:448.17188

Point_id＝227[73095.0]clusterId:1 dist:150.82812

Point_id＝231[71891.0]clusterId:1 dist:1053.1719

Point_id＝232[72563.0]clusterId:1 dist:381.17188

Point_id＝235[73865.0]clusterId:1 dist:920.8281

Point_id＝238[71708.0]clusterId:1 dist:1236.1719

Point_id＝239[71616.0]clusterId:1 dist:1328.1719

Point_id＝242[71864.0]clusterId:1 dist:1080.1719

Point_id＝243[71585.0]clusterId:1 dist:1359.1719

Point_id＝248[73503.0]clusterId:1 dist:558.8281

Point_id＝251[72101.0]clusterId:1 dist:843.1719

Point_id＝253[72701.0]clusterId:1 dist:243.17188

Point_id＝259[72083.0]clusterId:1 dist:861.1719

Point_id＝263[74147.0]clusterId:1 dist:1202.8281

Point_id＝266[73084.0]clusterId:1 dist:139.82812

Point_id＝270[72505.0]clusterId:1 dist:439.17188

Point_id＝274[73438.0]clusterId:1 dist:493.82812

Point_id＝277[73913.0]clusterId:1 dist:968.8281

Point_id＝280[73246.0]clusterId:1 dist:301.82812

Point_id＝283[72282.0]clusterId:1 dist:662.1719

Point_id＝284[71877.0]clusterId:1 dist:1067.1719

Point_id＝286[71797.0]clusterId:1 dist:1147.1719

Point_id＝288[72038.0]clusterId:1 dist:906.1719

Point_id＝295[72857.0]clusterId:1 dist:87.171875

Point_id＝297[71836.0]clusterId:1 dist:1108.1719

Point_id＝298[71528.0]clusterId:1 dist:1416.1719

Point_id＝300[71470.0]clusterId:1 dist:1474.1719

Point_id＝302[73814.0]clusterId:1 dist:869.8281

Point_id＝311[72361.0]clusterId:1 dist:583.1719

Point_id＝312[71951.0]clusterId:1 dist:993.1719

Point_id＝315[72606.0]clusterId:1 dist:338.17188

Point_id＝317[72324.0]clusterId:1 dist:620.1719

Point_id＝321[71768.0]clusterId:1 dist:1176.1719

dest_avg：769.47015

Cluster

Cluster_id＝4,center:{Point_id＝-1[17627.0]clusterId:0 dist:0.0}

Point_id＝12[10014.0]clusterId:4 dist:7613.0

Point_id＝63[13868.0]clusterId:4 dist:3759.0

Point_id＝107[21984.0]clusterId:4 dist:4357.0

Point_id＝167[6908.0]clusterId:4 dist:10719.0

Point_id＝269[26874.0]clusterId:4 dist:9247.0

Point_id＝307[26114.0]clusterId:4 dist:8487.0

dest_avg：7363.6665

Cluster

Cluster_id＝3,center:{Point_id＝-1[69303.64]clusterId:0 dist:0.0}

Point_id＝2[68113.0]clusterId:3 dist:1190.6406

Point_id＝3[69274.0]clusterId:3 dist:29.640625

Point_id＝4[68626.0]clusterId:3 dist:677.6406

Point_id＝19[70978.0]clusterId:3 dist:1674.3594

Point_id＝38[70992.0]clusterId:3 dist:1688.3594

Point_id＝96[69764.0]clusterId:3 dist:460.35938

Point_id＝140[71068.0]clusterId:3 dist:1764.3594

Point_id＝143[68741.0]clusterId:3 dist:562.6406

Point_id＝144[66867.0]clusterId:3 dist:2436.6406

Point_id＝152[71113.0]clusterId:3 dist:1809.3594

Point_id＝166[64379.0]clusterId:3 dist:4924.6406

Point_id＝174[68586.0]clusterId:3 dist:717.6406

Point_id＝192[70223.0]clusterId:3 dist:919.3594

Point_id＝196[65357.0]clusterId:3 dist:3946.6406

Point_id＝229[71099.0]clusterId:3 dist:1795.3594

Point_id＝244[67437.0]clusterId:3 dist:1866.6406

Point_id＝247[67895.0]clusterId:3 dist:1408.6406

Point_id＝254[70085.0]clusterId:3 dist:781.3594

Point_id＝257[71041.0]clusterId:3 dist:1737.3594

Point_id＝260[70710.0]clusterId:3 dist:1406.3594

Point_id＝265[69472.0]clusterId:3 dist:168.35938

Point_id＝273[68006.0]clusterId:3 dist:1297.6406

Point_id＝279[70714.0]clusterId:3 dist:1410.3594

Point_id＝282[68152.0]clusterId:3 dist:1151.6406

Point_id＝290[70633.0]clusterId:3 dist:1329.3594

Point_id＝304[70765.0]clusterId:3 dist:1461.3594

Point_id＝308[70323.0]clusterId:3 dist:1019.3594

Point_id＝319[70089.0]clusterId:3 dist:785.3594

dest_avg：1443.6228

Cluster

Cluster_id＝0,center:{Point_id＝-1[45550.848]clusterId:0 dist:0.0}

Point_id＝0[45026.0]clusterId:0 dist:524.84766

Point_id＝1[46579.0]clusterId:0 dist:1028.1523

Point_id＝5[43922.0]clusterId:0 dist:1628.8477

Point_id＝7[41586.0]clusterId:0 dist:3964.8477

Point_id＝8[43835.0]clusterId:0 dist:1715.8477

Point_id＝11[43391.0]clusterId:0 dist:2159.8477

Point_id＝15[57407.0]clusterId:0 dist:11856.152

Point_id＝18[43825.0]clusterId:0 dist:1725.8477

Point_id＝20[45638.0]clusterId:0 dist:87.15234

Point_id＝22[43444.0]clusterId:0 dist:2106.8477

Point_id＝25[45514.0]clusterId:0 dist:36.847656

Point_id＝26[43944.0]clusterId:0 dist:1606.8477

Point_id＝28[44890.0]clusterId:0 dist:660.84766

Point_id＝30[55057.0]clusterId:0 dist:9506.152

Point_id＝32[43251.0]clusterId:0 dist:2299.8477

Point_id＝34[42857.0]clusterId:0 dist:2693.8477

Point_id＝35[42611.0]clusterId:0 dist:2939.8477

Point_id＝37[45313.0]clusterId:0 dist:237.84766

Point_id＝39[39994.0]clusterId:0 dist:5556.8477

Point_id＝40[43252.0]clusterId:0 dist:2298.8477

Point_id＝42[45051.0]clusterId:0 dist:499.84766

Point_id＝44[43889.0]clusterId:0 dist:1661.8477

Point_id＝46[44850.0]clusterId:0 dist:700.84766

Point_id＝48[43532.0]clusterId:0 dist:2018.8477

Point_id＝51[45725.0]clusterId:0 dist:174.15234

Point_id＝53[43425.0]clusterId:0 dist:2125.8477

Point_id＝54[45132.0]clusterId:0 dist:418.84766

Point_id＝57[46208.0]clusterId:0 dist:657.15234

Point_id＝58[45834.0]clusterId:0 dist:283.15234

Point_id＝62[45655.0]clusterId:0 dist:104.15234

Point_id＝64[45091.0]clusterId:0 dist:459.84766

Point_id＝65[41964.0]clusterId:0 dist:3586.8477

Point_id＝67[47841.0]clusterId:0 dist:2290.1523

Point_id＝70[46028.0]clusterId:0 dist:477.15234

Point_id＝72[42944.0]clusterId:0 dist:2606.8477

Point_id＝74[45131.0]clusterId:0 dist:419.84766

Point_id＝76[48199.0]clusterId:0 dist:2648.1523

Point_id＝79[48581.0]clusterId:0 dist:3030.1523

Point_id＝80[48055.0]clusterId:0 dist:2504.1523

Point_id＝83[44899.0]clusterId:0 dist:651.84766

Point_id＝86[44955.0]clusterId:0 dist:595.84766

Point_id＝89[46798.0]clusterId:0 dist:1247.1523

Point_id＝90[47202.0]clusterId:0 dist:1651.1523

Point_id＝93[44871.0]clusterId:0 dist:679.84766

Point_id＝95[43923.0]clusterId:0 dist:1627.8477

Point_id＝98[46641.0]clusterId:0 dist:1090.1523

Point_id＝99[44380.0]clusterId:0 dist:1170.8477

Point_id＝101[46171.0]clusterId:0 dist:620.15234

Point_id＝104[46169.0]clusterId:0 dist:618.15234

Point_id＝106[45103.0]clusterId:0 dist:447.84766

Point_id＝108[44052.0]clusterId:0 dist:1498.8477

Point_id＝109[47174.0]clusterId:0 dist:1623.1523

Point_id＝111[46942.0]clusterId:0 dist:1391.1523

Point_id＝114[45056.0]clusterId:0 dist:494.84766

Point_id＝116[45288.0]clusterId:0 dist:262.84766

Point_id＝117[45994.0]clusterId:0 dist:443.15234

Point_id＝120[46563.0]clusterId:0 dist:1012.15234

Point_id＝122[46124.0]clusterId:0 dist:573.15234

Point_id＝124[44983.0]clusterId:0 dist:567.84766

Point_id＝126[44941.0]clusterId:0 dist:609.84766

Point_id＝127[44736.0]clusterId:0 dist:814.84766

Point_id＝130[46836.0]clusterId:0 dist:1285.1523

Point_id＝132[45110.0]clusterId:0 dist:440.84766

Point_id＝134[41763.0]clusterId:0 dist:3787.8477

Point_id＝137[44544.0]clusterId:0 dist:1006.84766

Point_id＝139[46786.0]clusterId:0 dist:1235.1523

Point_id＝141[48905.0]clusterId:0 dist:3354.1523

Point_id＝145[44763.0]clusterId:0 dist:787.84766

Point_id＝147[44327.0]clusterId:0 dist:1223.8477

Point_id＝149[44368.0]clusterId:0 dist:1182.8477

Point_id＝150[46410.0]clusterId:0 dist:859.15234

Point_id＝153[46643.0]clusterId:0 dist:1092.1523

Point_id＝154[42034.0]clusterId:0 dist:3516.8477

Point_id＝156[46444.0]clusterId:0 dist:893.15234

Point_id＝158[47361.0]clusterId:0 dist:1810.1523

Point_id＝159[46867.0]clusterId:0 dist:1316.1523

Point_id＝161[49011.0]clusterId:0 dist:3460.1523

Point_id＝164[42106.0]clusterId:0 dist:3444.8477

Point_id＝165[46624.0]clusterId:0 dist:1073.1523

Point_id＝169[48455.0]clusterId:0 dist:2904.1523

Point_id＝172[52153.0]clusterId:0 dist:6602.1523

Point_id＝173[45033.0]clusterId:0 dist:517.84766

Point_id＝177[46142.0]clusterId:0 dist:591.15234

Point_id＝179[45843.0]clusterId:0 dist:292.15234

Point_id＝180[44687.0]clusterId:0 dist:863.84766

Point_id＝182[42359.0]clusterId:0 dist:3191.8477

Point_id＝185[44943.0]clusterId:0 dist:607.84766

Point_id＝187[44881.0]clusterId:0 dist:669.84766

Point_id＝188[46526.0]clusterId:0 dist:975.15234

Point_id＝191[44498.0]clusterId:0 dist:1052.8477

Point_id＝193[45735.0]clusterId:0 dist:184.15234

Point_id＝197[45581.0]clusterId:0 dist:30.152344

Point_id＝200[43951.0]clusterId:0 dist:1599.8477

Point_id＝202[45406.0]clusterId:0 dist:144.84766

Point_id＝203[42156.0]clusterId:0 dist:3394.8477

Point_id＝204[45209.0]clusterId:0 dist:341.84766

Point_id＝207[47754.0]clusterId:0 dist:2203.1523

Point_id＝209[44713.0]clusterId:0 dist:837.84766

Point_id＝211[47567.0]clusterId:0 dist:2016.1523

Point_id＝212[46270.0]clusterId:0 dist:719.15234

Point_id＝215[44731.0]clusterId:0 dist:819.84766

Point_id＝216[45345.0]clusterId:0 dist:205.84766

Point_id＝219[45353.0]clusterId:0 dist:197.84766

Point_id＝220[45837.0]clusterId:0 dist:286.15234

Point_id＝222[44873.0]clusterId:0 dist:677.84766

Point_id＝224[41313.0]clusterId:0 dist:4237.8477

Point_id＝226[46166.0]clusterId:0 dist:615.15234

Point_id＝228[44966.0]clusterId:0 dist:584.84766

Point_id＝230[48399.0]clusterId:0 dist:2848.1523

Point_id＝233[44865.0]clusterId:0 dist:685.84766

Point_id＝234[46014.0]clusterId:0 dist:463.15234

Point_id＝237[46294.0]clusterId:0 dist:743.15234

Point_id＝240[44893.0]clusterId:0 dist:657.84766

Point_id＝241[46607.0]clusterId:0 dist:1056.1523

Point_id＝245[45607.0]clusterId:0 dist:56.152344

Point_id＝246[45471.0]clusterId:0 dist:79.84766

Point_id＝249[44394.0]clusterId:0 dist:1156.8477

Point_id＝250[45699.0]clusterId:0 dist:148.15234

Point_id＝252[44071.0]clusterId:0 dist:1479.8477

Point_id＝255[44541.0]clusterId:0 dist:1009.84766

Point_id＝256[49629.0]clusterId:0 dist:4078.1523

Point_id＝258[48557.0]clusterId:0 dist:3006.1523

Point_id＝261[46739.0]clusterId:0 dist:1188.1523

Point_id＝262[45764.0]clusterId:0 dist:213.15234

Point_id＝264[45937.0]clusterId:0 dist:386.15234

Point_id＝267[47049.0]clusterId:0 dist:1498.1523

Point_id＝268[45604.0]clusterId:0 dist:53.152344

Point_id＝271[44946.0]clusterId:0 dist:604.84766

Point_id＝272[44858.0]clusterId:0 dist:692.84766

Point_id＝276[45329.0]clusterId:0 dist:221.84766

Point_id＝278[45673.0]clusterId:0 dist:122.15234

Point_id＝281[45086.0]clusterId:0 dist:464.84766

Point_id＝285[46088.0]clusterId:0 dist:537.15234

Point_id＝287[47575.0]clusterId:0 dist:2024.1523

Point_id＝289[43379.0]clusterId:0 dist:2171.8477

Point_id＝291[44109.0]clusterId:0 dist:1441.8477

Point_id＝292[48403.0]clusterId:0 dist:2852.1523

Point_id＝294[43930.0]clusterId:0 dist:1620.8477

Point_id＝296[48220.0]clusterId:0 dist:2669.1523

Point_id＝299[44775.0]clusterId:0 dist:775.84766

Point_id＝301[45165.0]clusterId:0 dist:385.84766

Point_id＝303[43404.0]clusterId:0 dist:2146.8477

Point_id＝305[46562.0]clusterId:0 dist:1011.15234

Point_id＝306[45553.0]clusterId:0 dist:2.1523438

Point_id＝309[44993.0]clusterId:0 dist:557.84766

Point_id＝310[44857.0]clusterId:0 dist:693.84766

Point_id＝313[46843.0]clusterId:0 dist:1292.1523

Point_id＝314[45947.0]clusterId:0 dist:396.15234

Point_id＝316[46538.0]clusterId:0 dist:987.15234

Point_id＝318[46515.0]clusterId:0 dist:964.15234

Point_id＝320[46512.0]clusterId:0 dist:961.15234

dest_avg：1434.0354

Cluster

Cluster_id＝2,center:{Point_id＝-1[75850.0]clusterId:0 dist:0.0}

Point_id＝16[75605.0]clusterId:2 dist:245.0

Point_id＝27[74668.0]clusterId:2 dist:1182.0

Point_id＝31[76935.0]clusterId:2 dist:1085.0

Point_id＝41[74455.0]clusterId:2 dist:1395.0

Point_id＝61[74796.0]clusterId:2 dist:1054.0

Point_id＝68[75231.0]clusterId:2 dist:619.0

Point_id＝69[77323.0]clusterId:2 dist:1473.0

Point_id＝77[76443.0]clusterId:2 dist:593.0

Point_id＝81[75978.0]clusterId:2 dist:128.0

Point_id＝94[76131.0]clusterId:2 dist:281.0

Point_id＝103[75742.0]clusterId:2 dist:108.0

Point_id＝105[79219.0]clusterId:2 dist:3369.0

Point_id＝133[75515.0]clusterId:2 dist:335.0

Point_id＝142[76980.0]clusterId:2 dist:1130.0

Point_id＝190[75455.0]clusterId:2 dist:395.0

Point_id＝206[74711.0]clusterId:2 dist:1139.0

Point_id＝210[74559.0]clusterId:2 dist:1291.0

Point_id＝236[75554.0]clusterId:2 dist:296.0

dest_avg：895.44446

on the basis of the analysis, constraint conditions are added (condition 1: the cluster radius needs to be 3600, namely the span of a time period is not more than 2 hours, condition 2: the data example contained in each cluster is not less than the total data amount 10%), the cluster meeting the constraint conditions is judged to be an effective cluster, so that the effective cluster of the case is 2 (20: 02:54-20:28:33, 12:15:16-13:03:04 every day), the ONU alarms in the time period can be identified as fault alarms caused by suspected customer outage, a notification work order is dispatched, the alarms not in the time period can be identified as alarms caused by other faults, and the fault work order needs to be dispatched.

Those skilled in the art will appreciate that the invention may be practiced without these specific details. It is pointed out that the above description is helpful for the person skilled in the art to understand the invention, but does not limit the scope of the invention. Any such equivalents, modifications and/or omissions as may be made without departing from the spirit and scope of the invention may be resorted to.

Claims

1. An ONU client side power failure intelligent judgment method based on machine learning is characterized in that according to the phenomenon that ONU reports an alarm caused by the fact that a client self-power failure exists in a fixed time interval in ONU operation and maintenance, a K-means clustering algorithm is utilized, a model for detecting the ONU client side power failure is obtained through analysis, whether the ONU power failure alarm is suspected to be caused by the client side power failure is judged in advance, and direct marking is carried out for classification processing, so that the link that maintenance personnel manually dial a call to contact with the client to confirm whether the client self-power failure alarm is caused can be omitted, the labor cost is reduced, the fault processing waiting time is shortened, the ONU power failure fault work order quantity is reduced, and the fault work order distribution accuracy and the processing efficiency are effectively improved.

2. The ONU client-side power-down intelligent judgment method based on machine learning of claim 1, comprising the following steps:

step 1, collecting ONU power-down alarm;

step 2, carrying out resource positioning aiming at ONU power failure alarm;

step 5a, marking a suspected client side power-off mark;

step 6a, dispatching a notification work order;

step 5b, undoubtedly, a client side power-off mark is similar;

step 6b, dispatching a fault work order;

step 7b, the maintenance personnel receive orders and process faults, and wait for receiving a clearing alarm;

step 8, returning the order;

and 9, archiving.

3. The method according to claim 2, wherein the ONU power-down alarm in step 1 comprises one or more of the following: the branch optical fiber is broken or the OLT cannot detect the expected optical signal LOSi/LOBi of the ONT; ONU in GPON alarm is disconnected; ONU signals in GPON alarm are invalid; ONU signal loss in GPON alarm; LINK _ LOSS in LINK alarm.

4. The ONU client side power failure intelligent judgment method based on machine learning according to claim 2, wherein the step 2 comprises that an operator fault center obtains resource data through a resource management system interface, whether equipment attribution client information and area information matched with ONU power failure alarm are complete or not is positioned, if not, the flow is ended, and if so, the step 3 is entered.

5. The ONU client side power failure intelligent judgment method based on machine learning of claim 2, wherein the step 3 comprises the steps of transmitting the ONU name, the home client information, the home area information and the alarm occurrence time of the ONU power failure alarm into a K-means machine learning model, detecting and identifying alarm data by the model, and marking a suspected client side power failure mark for the alarm which has consistent resources and meets the suspected client side power failure time period.

6. The method for intelligently judging the power failure of the ONU client side based on machine learning according to claim 2, wherein said step 3 comprises the following steps:

step K1, collecting ONU power-down history alarm;

k2, performing sampling analysis on ONU power-down history alarms;

7. The method according to claim 6, wherein the step K1 includes extracting the resource, occurrence time and recovery time of the ONU power-down alarm from the fault center history alarm library, extracting the fault cause of the fault work order related to the alarm, and analyzing that the number of the ONU samples with the assignment amount exceeding 50 exceeds 160, or the number of the ONU samples with the assignment amount exceeding 20 exceeds 1000.

8. The ONU client-side power failure intelligent judgment method based on machine learning of claim 6, wherein the step K2 comprises performing scatter plot processing on the occurrence time and recovery time of the corresponding alarm of which the failure cause is the client-side power failure to find out the regularity of the occurrence time and recovery time of part of the ONU power failure alarms.

9. The ONU client-side power-off intelligent judgment method based on machine learning of claim 6, wherein the K-means clustering algorithm in the step K3 adopts the following algorithm formula:

10. The ONU client-side power-off intelligent judgment method based on machine learning of claim 1, wherein in the step K4, each cluster has a center point parameter and a cluster radius parameter; the step K5 includes analyzing the ONU with the highest dispatching frequency every day through the continuously accumulated training sample data, excavating and updating the time period of the suspected user power failure habit, comparing the real-time alarm with the time period, dispatching the alarm meeting the time period to notify the work order, and dispatching the alarm not meeting the time period to send the fault work order.