WO2023216457A1

WO2023216457A1 - Method for predicting and positioning abnormity of transmission network between core network and base station

Info

Publication number: WO2023216457A1
Application number: PCT/CN2022/114151
Authority: WO
Inventors: 王玉梁; 朱文进
Original assignee: 中电信数智科技有限公司
Priority date: 2022-05-11
Filing date: 2022-08-23
Publication date: 2023-11-16
Also published as: CN115022908B; CN115022908A

Abstract

A method for predicting and positioning an abnormity of a transmission network between a core network and a base station. The method comprises: classifying end-to-end troubleshooting process throughput data of a transmission network between the core network and a gNB, wherein the end-to-end process comprises five links, i.e., a server, the core network, a transmission link, the gNB, and UE as well as a PC connected to the UE; analyzing historical log data of each link to obtain an initial fault occurrence probability; moreover obtaining real-time data by means of current inspection and troubleshooting, and scoring each link according to the throughput rate; analyzing historical and real-time data to obtain real-time data that a fault is transferred to normality and the normality is transferred to a fault, and generating a rectangular set; then putting the rectangular set composed of three sets of data into an artificial intelligence model to analyze to obtain a future fault occurrence probability of each link; and generating a fault probability and scoring baseline by scoring the throughput during troubleshooting of each link in combination with the future fault occurrence probability of the link, comparing and analyzing, and troubleshooting and positioning a fault by means of an analysis result.

Description

A method for predicting and locating abnormality in core network and base station transmission network

Technical field

The present invention relates to the technical field of network troubleshooting, and in particular to a method for predicting and locating abnormality in a core network and base station transmission network.

Background technique

In the network communication environment, the structure and environment of the transmission network between the core network and gNB (base station) are complex; therefore, when an abnormality occurs in the network, it is easy for the operation and maintenance personnel to have insufficient traceability and tracking capabilities for the abnormal points, and the connection between the networks Tight, some hidden fault points are difficult to troubleshoot.

This application divides the end-to-end troubleshooting process of the transmission network between the core network and gNB into five links: server, core network, transmission link, gNB, UE, and PC connected to the UE. By comparing and analyzing the throughput score of each link during troubleshooting and the future failure probability of the link, the failure probability is generated and compared with the scoring baseline to troubleshoot and locate the fault.

Contents of the invention

In view of the shortcomings in the existing technology, the present invention provides a method for predicting and locating abnormality in the core network and base station transmission network; using artificial intelligence analysis and prediction, various network fault (abnormal) characteristics can be quickly and timely checked and positioned quickly. Fault link. It can better help maintenance personnel discover hidden faults in network nodes, restore network communications as soon as possible, and improve user experience.

In order to achieve the above objects, the present invention adopts the following technical solutions:

A method for predicting and locating abnormality in core network and base station transmission network, including the following steps:

S1: Make statistics on each node in the transmission network between the core network and the base station, classify the secondary indicators according to the indicator type of each node; regularly evaluate and score each node according to the status of the secondary indicator classification, and use this calculation Score for each secondary indicator category;

S2: When conducting the nth round of evaluation process, obtain the historical data before the nth round of evaluation, calculate the abnormality probability of each secondary indicator classification, and use it as the initial abnormality probability a of the nth round of evaluation; calculate each The normal occurrence probability of the secondary indicator classification is used as the initial normal occurrence probability b of the nth round of evaluation;

S3: Obtain the scores of each secondary indicator category in the current evaluation round: nth round of evaluation, and compare them with the scores of the corresponding secondary indicator categories in the historical data, and then determine the performance of each secondary indicator category in the next round n+ 1. The probability of transitioning from abnormality to normal during the evaluation process c. The probability that each secondary indicator classification will change from abnormal to remain abnormal in the next n+1 evaluation process d. The probability that each secondary indicator classification will remain abnormal during the next n+1 evaluation process. The probability of transitioning from normal to abnormal is e, and the probability of each secondary indicator classification changing from normal to remaining normal in the next n+1 evaluation process f;

S4: Based on the current evaluation round: the initial abnormal occurrence probability a of the nth round of evaluation, and the initial normal occurrence probability b of the current evaluation round nth round of evaluation, predict and calculate the abnormality of each secondary indicator classification in the n+1 round of evaluation process Occurrence probability g and normal occurrence probability h;

S5: When the n+1th round of evaluation is performed, the current evaluation round n=n+1 at this time, and the abnormality occurrence probability g predicted and calculated in step S4 is used as the initial abnormality occurrence probability a in the evaluation process of this round, and normal The occurrence probability h is used as the initial normal occurrence probability b during this round of evaluation;

S6: Loop steps S3-S5, that is, the data of each evaluation and its historical data can be combined to calculate the abnormal occurrence probability and normal occurrence probability of each secondary indicator classification in the subsequent round of evaluation, so as to conduct a comparison between the core network and Predict and locate abnormal conditions of each node in the transmission network between base stations.

In order to optimize the above technical solutions, specific measures taken also include:

Further, the specific content of step S1 is:

S1.1: Traverse all nodes in the transmission network between the core network and the base station; including: server, core network, transmission link, gNB, UE and PC connected to the UE;

S1.2: Classify secondary indicators according to the indicator type of each node; the server includes 2 secondary indicator classifications: hardware performance status and parameter setting status; the core network includes 2 secondary indicator classifications: core network Speed limit status, bearer network status; the transmission link includes three secondary indicator categories: transmission link bandwidth limit status, transmission link large delay jitter status, and transmission link packet loss and disorder status; the gNB includes 3 secondary indicator categories: base station rate limit status, base station processing capability status, algorithm feature limit status; the UE and the PC connected to the UE include terminal capability status, PC performance status, TCP setting status, and software setting status;

S1.3: Set the weight of each secondary indicator classification in the node score of the corresponding link;

S1.4: Regularly score the corresponding link nodes according to the status of the secondary indicator classification; at the same time, obtain the score of each secondary indicator classification in each round of evaluation based on the score proportion weight.

Further, the specific content of step S2 is: determining the overall number of evaluations before the nth round of evaluation through historical data, and determining the number of abnormalities in each secondary index category in the overall number of evaluations, and calculating the number of abnormalities in each secondary index category. The proportion of the number of abnormal occurrences in the total number of evaluations determines the initial abnormal occurrence probability a of the corresponding secondary indicator classification in the nth round of evaluation. Therefore, the initial normal occurrence probability b=1 of the corresponding secondary indicator classification in the nth round of evaluation. -a.

Further, the specific content of step S3 is: obtain the scores of each secondary index category in the historical data before the current nth round of evaluation, and perform a weighted average to obtain the average score of each secondary index category, and add the scores in the current nth round of evaluation to The score of each secondary indicator category is compared with the average score of the corresponding secondary indicator category to determine the increase or decrease range. Based on the increase or decrease range, the probability that each secondary indicator category will transition from abnormal to normal in subsequent evaluations is determined. c, and calculate the probability d=1-c that each secondary indicator classification changes from abnormal to remain abnormal in subsequent evaluations; similarly, determine whether each secondary indicator classification changes from normal to normal in subsequent evaluations based on the range of increase or decrease. The probability of abnormality e, and calculate the probability f=1-c that each secondary indicator classification changes from normal to remain normal in subsequent evaluations.

Further, the specific content of step S4 is: predicting and calculating the abnormality occurrence probability g and normal occurrence probability h of each secondary index classification of the n+1th round of evaluation and the initial abnormality occurrence probability a, initial abnormality occurrence probability a of the nth round of the current evaluation round. The relationship between the normal occurrence probabilities b is: g=a*d+b*e, h=a*c+b*f.

Furthermore, it also includes,

Step S7: Select each secondary indicator prediction data and evaluation score data obtained from some rounds of evaluation in the multi-round evaluation process in step S6, generate a failure probability baseline and a score baseline, and conduct comparative analysis;

If the predicted abnormality probability increases and the score decreases, the two results are consistent and it is normal;

If the predicted abnormality probability decreases and the score increases, the two results are consistent and it is normal;

If the predicted abnormality probability increases and the score decreases, the two results do not match, which is an abnormal situation, indicating that there are hidden abnormalities in the corresponding secondary indicator classification;

If the predicted abnormality probability decreases and the score increases, the two results do not match, which is an abnormal situation, indicating that there are hidden abnormalities in the secondary indicator classification;

Check the secondary indicator classification based on the matching status of the two.

The beneficial effects of the present invention are:

1. This application combines the current round of evaluation data and previous historical data to calculate and predict the abnormal occurrence probability and normal occurrence probability of each secondary indicator classification in the subsequent round of evaluation, which can be used as reference data for operation and maintenance personnel. Operation and maintenance personnel can focus on troubleshooting secondary classification indicators with a high probability of abnormality to reduce troubleshooting time for abnormal fault points. At the same time, this application also helps to query some hidden and difficult-to-diagnose fault points through high-risk secondary classification indicators; it also provides operation and maintenance personnel with more reliable troubleshooting suggestions.

2. This application uses the prediction data of each secondary indicator and the evaluation scoring data in the multiple rounds of evaluation processes to generate a fault probability baseline and a scoring baseline, and conducts comparative analysis; the two are mutually restricted and compared to determine whether there are hidden faults. exception to help operation and maintenance personnel troubleshoot.

3. This application highlights the role of artificial intelligence in optimizing 5G cell core network and base station transmission network fault prediction. The use of artificial intelligence analysis and prediction can quickly and timely investigate various network fault characteristics and quickly locate fault links. It can better help maintenance personnel discover hidden faults in network nodes, restore network communications as soon as possible, and improve user experience.

Description of the drawings

Figure 1 is a schematic diagram of the connection relationship of each link node in the network transmission process of the present invention and the index relationship mainly involved in the corresponding node.

Figure 2 is a schematic diagram for calculating the relationship between the data predicted by the next round of evaluation and the initial abnormality/normal occurrence probability in the embodiment of the present invention.

Figure 3 is a schematic diagram comparing the failure probability baseline and the scoring baseline in the embodiment of the present invention.

Detailed ways

The present invention will now be described in further detail with reference to the accompanying drawings.

Referring to Figures 1-3, the main technical solutions of this application are:

Step 1: First, classify the possible failure factors and related indicators corresponding to each link in the troubleshooting process of the transmission network between the core network and gNB. The end-to-end process includes five links (server, core network, transmission link, gNB, UE and PC connected to the UE). The sum of the scores for each link is 1, among which the weighted average distribution of the score indicators is based on the number of secondary categories. The initial failure probability is obtained through the analysis of historical log data of each link; at the same time, real-time data is obtained during this inspection, and each link is scored according to the reference indicators of each link in the process; the normal and real-time failover is obtained through historical and real-time data analysis Data transferred normally to failure generates a rectangular collection.

The first category and server link are divided into 2 level 2 categories:

Secondary classification of hardware performance and parameter settings:

1. Hardware performance

Check whether there are hardware alarms (optical module, disk damage, memory module abnormality, board failure, CPU temperature is too high, memory temperature is too high) according to the log files stored on the transmission network equipment between the core network and gNB. These are all hardware problems. .

2. Parameter settings

Query whether there are TCP and software parameter alarms (TCP parameter alarms, software parameter alarms) based on the log files stored on the transmission network equipment between the core network and gNB, which are all parameter setting issues. TCP parameter alarms (SYN Cookies, fast recycling of sockets, port range for outgoing connections, length of SYN queue, maximum number of TIME_WAITs maintained by the system at the same time), software parameter alarms (maximum number of database connections, maximum memory access settings).

The second category and core network links are divided into two level 2 categories:

Speed-limited, dedicated bearer secondary classification:

1. Core network speed limit

Check whether the ratio of the set speed limit and bandwidth configuration parameters is correct based on the configuration file stored on the core network device. For example: if the bandwidth is configured as 1000M and the speed limit is 500, then the maximum bandwidth utilization is only 50% and needs to be adjusted.

2. Bearer network

First, check the account opening information through base station side signaling tracking to see if the core network has other special configurations. If so, query whether there is a bearer network and corresponding alarm data based on the log files stored on the transmission network equipment between the core network and gNB. Bearer network: It is to open the connection between the base station and the central computer room. The network responsible for carrying and aggregating data.

The third category, transmission link links are divided into three 2-level classifications:

Transmission link bandwidth limitations, large delay jitter, and packet loss out of order:

1. Transmission link bandwidth limitation

Check whether the transmission bandwidth is limited according to the configuration file stored on the gNB.

2. Large delay jitter in transmission links

Collect the network traffic on the transmission link and compare whether there is a sudden increase or decrease in delay jitter between the last inspection and this inspection.

3. Packet loss and disorder in transmission link

Collect the network traffic on the transmission link and combine it with the alarm information to check whether there are packet loss and out-of-sequence alarms based on UDP and TCP protocols, and implement packet loss retransmission mechanisms and timeout mechanisms for UDP and TCP protocols.

The fourth category, gNB link, is divided into three 2-level classifications:

Base station rate limitations, base station processing capabilities, and algorithm feature limitations:

1. Base station rate limit

Check whether the ratio of the set speed limit and bandwidth configuration parameters is correct according to the configuration file stored on the gNB. For example: if the bandwidth is configured as 1000M and the speed limit is 500, then the maximum bandwidth utilization is only 50% and needs to be adjusted.

2. Base station processing capability

Query based on the configuration file stored on gNB:

S1, whether the basic configuration such as bandwidth and number of antennas is unreasonable and causes throughput problems;

S2. Replace the test equipment and confirm whether it is a test terminal problem;

S3. Poor wireless signal will inevitably lead to reduced throughput, so we test the wireless signal quality at two locations near and far within the community.

Caused by near points, abnormal far points: cell edge interference causes throughput problems, find the source of interference, and perform RF optimization. RF optimization includes four stages: preparation, data collection, problem analysis, and adjustment and implementation. Data collection, problem analysis, and optimization adjustment need to be repeated according to the optimization target requirements and the actual optimization status, until the network conditions meet the optimization target KPI requirements. .

The near point is abnormal, the far point is normal: Check the terminal capability to determine whether the current rate is close to the theoretical peak.

3. Limitations of algorithm characteristics

Check whether the encryption algorithm and storage algorithm configuration are correct according to the configuration file stored on the gNB.

The fifth category, UE and PC links connected to UE are divided into 4 level 2 categories:

Terminal capabilities, PC performance, TCP settings, software configuration (FTP configuration, firewall);

1. Terminal capabilities

Check whether the UE terminal configuration is too low through the measurement report uploaded by the customer.

2. PC performance

Check whether the performance of the service PC connected to the UE meets the requirements and whether the configuration on the UE side is correct.

3. TCP settings

Check whether the scheduling is sufficient under the premise that the data source is sufficient and the channel conditions are good. Check whether the channel quality meets the requirements (the peak test requires SINR > 25dB and the block error rate is 0).

4. Software configuration

FTP configuration and firewall configuration.

Among them, in one embodiment, the investigation results are as shown in Table 1 below:

Table 1 Transmission network troubleshooting results between the core network and gNB

Step 2: Based on the [Fault Prediction Model] analysis, the probability of failure in each link of the transmission network throughput between the core network and gNB is obtained next time. Finally, the failure probability of each link is generated to generate a fault probability baseline, and the scores are used to generate a scoring baseline.

S1. Build [Fault Prediction Model]

[Fault prediction model] formula:

Probability matrix model formula: X(k+1)=X(k)×P

In the formula: X(k) represents the state vector of the trend analysis and prediction object at time t=k, P represents the one-step transition probability matrix, and vector.

S1-1. Analyze the secondary classification keywords by accessing the data sources corresponding to the secondary classification of the current troubleshooting step (configuration files stored on gNB or logs stored on the transmission network equipment between the core network and gNB), and obtain Historical data generates initial probability occurrence data;

For example: Program troubleshooting -> Server link -> Hardware performance secondary classification obtains a failure probability of 30% through historical data, then the probability of normal occurrence is 1-30% = 0.7 [0.3 0.7].

S1-2. Compare the data collected in this investigation (memory, cpu, disk size, etc.) with the weighted average data of the S1 data source historical monitoring indicators (memory, cpu, disk size, etc.) to obtain the increase or decrease. That is, the normal proportion of hardware performance failover during this troubleshooting process;

For example: Program troubleshooting -> Server link -> Hardware performance secondary classification. By comparing the data collected during this troubleshooting (memory, cpu, disk size, etc.) with historical data, we can obtain 40% of the normal proportion of hardware performance failover during this troubleshooting process. , then the proportion of faults remaining faulty is 1-0.4=0.6 [0.4 0.6].

(The transition probability is specifically determined based on the increase or decrease range. For example, the hardware performance in the secondary indicator classification of the server has a score of a in this evaluation, and the weighted average in historical evaluations is Score b, the increase or decrease range is c; when c is located in a certain interval, set the probability of normal or normal failover to d%; when c is located in another interval, set its failover normal or normal The probability of transfer failure is e%, and this is used to determine the transfer probability)

S1-3. Compare the data collected in this investigation (memory, cpu, disk size, etc.) with the weighted average data of the S1 data source historical monitoring indicators (memory, cpu, disk size, etc.) to obtain the increase or decrease. That is, the proportion of hardware performance normal transfer failures during this troubleshooting process;

For example: Program troubleshooting -> Server link -> Hardware performance secondary classification. By comparing the collected data (memory, cpu, disk size) of this troubleshooting with historical data, we can obtain that the proportion of normal transfer failures in hardware performance during this troubleshooting process is 30%. Then the ratio of normal still remaining normal is 1-0.3=0.7 [0.3 0.7].

Referring to Figure 2, perform rectangular aggregation of the three sets of data obtained by the operation, and perform the operation to obtain the result, then:

Next time - the calculation process and results of the probability of occurrence of secondary classification faults in this link:

1. Failure probability = 0.3x0.6+0.3x0.7=0.39

2. Failure probability = 0.3x0.4+7x0.7=0.61

Server link -> Hardware performance secondary classification: The probability of troubleshooting next time is 39%, and the probability of non-occurrence is 61% [0.39 0.61]

Taking [0.39 0.61] data as the initial probability, assuming that the transition probabilities of S1-2 and S1-3 do not change, the next failure probability will occur.

1. Failure probability = 0.39x0.6+0.61x0.3=0.417

2. Failure probability = 0.39x0.4+0.61x0.7=0.583

Server link -> Hardware performance secondary classification: The probability of troubleshooting next time is 41.7%, and the probability of non-occurrence is 58.3%. 【0.417 0.583】

Step 3: Compare and analyze the failure probability with the scoring baseline to troubleshoot and locate the failure. Abbreviation of failure probability baseline: (A), Abbreviation of scoring baseline: (B) Comparison formula:

A rises and B falls: the probability of failure becomes greater and the score becomes lower (normal) (refer to the trend chart on the right side of Figure 3)

A falls and B rises: the probability of failure becomes smaller and the score becomes higher (normal)

A rises and B rises: The probability of failure becomes greater and the score becomes higher (abnormal) (refer to the trend chart on the left in Figure 3)

Detailed description: The failure probability increases and the score is very high. The possible reasons are due to abnormal network jitter, sensitive false alarms, and hidden faults of other network nodes. Therefore, when A rises and B rises, focus on the first-level link combined with the alarm information.

A decrease and B decrease: the probability of failure becomes smaller and the score becomes lower (abnormal)

Detailed description: Possible reasons for the failure probability becoming smaller and the score being very low are, for example, occurring during peak business hours. Monitoring indicator values are too high but the time period falls within the peak period. You can appropriately adjust the alarm threshold of monitoring indicators during peak hours. If there are off-peak periods, the possible reason is to perform data backup when the network load is small, and at the same time, focus on investigating abnormal non-backup data based on real-time data from each link.

The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above-mentioned embodiments. All technical solutions that fall under the idea of the present invention belong to the protection scope of the present invention. It should be pointed out that for those of ordinary skill in the art, several improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

A method for predicting and locating abnormality in core network and base station transmission network, which is characterized by including the following steps:

S1: Make statistics on each node in the transmission network between the core network and the base station, classify the secondary indicators according to the indicator type of each node; regularly evaluate and score each node according to the status of the secondary indicator classification, and use this calculation Score for each secondary indicator category;

S2: When conducting the nth round of evaluation process, obtain the historical data before the nth round of evaluation, calculate the abnormality probability of each secondary indicator classification, and use it as the initial abnormality probability a of the nth round of evaluation; calculate each The normal occurrence probability of the secondary indicator classification is used as the initial normal occurrence probability b of the nth round of evaluation;

S3: Obtain the scores of each secondary indicator category in the current evaluation round: nth round of evaluation, and compare them with the scores of the corresponding secondary indicator categories in the historical data, and then determine the performance of each secondary indicator category in the next round n+ 1. The probability of transitioning from abnormality to normal during the evaluation process c. The probability that each secondary indicator classification will change from abnormal to remain abnormal in the next n+1 evaluation process d. The probability that each secondary indicator classification will remain abnormal during the next n+1 evaluation process. The probability of transitioning from normal to abnormal is e, and the probability of each secondary indicator classification changing from normal to remaining normal in the next n+1 evaluation process f;

S4: Based on the current evaluation round: the initial abnormal occurrence probability a of the nth round of evaluation, and the initial normal occurrence probability b of the current evaluation round nth round of evaluation, predict and calculate the abnormality of each secondary indicator classification in the n+1 round of evaluation process Occurrence probability g and normal occurrence probability h;

S5: When the n+1th round of evaluation is performed, the current evaluation round n=n+1 at this time, and the abnormality occurrence probability g predicted and calculated in step S4 is used as the initial abnormality occurrence probability a in the evaluation process of this round, and normal The occurrence probability h is used as the initial normal occurrence probability b during this round of evaluation;

S6: Loop steps S3-S5, that is, the data of each evaluation and its historical data can be combined to calculate the abnormal occurrence probability and normal occurrence probability of each secondary indicator classification in the subsequent round of evaluation, so as to conduct a comparison between the core network and Predict and locate abnormal conditions of each node in the transmission network between base stations.
A method for predicting and locating abnormality in core network and base station transmission network according to claim 1, characterized in that the specific content of step S1 is:

S1.1: Traverse all nodes in the transmission network between the core network and the base station; including: server, core network, transmission link, gNB, UE and PC connected to the UE;

S1.2: Classify secondary indicators according to the indicator type of each node; the server includes 2 secondary indicator classifications: hardware performance status and parameter setting status; the core network includes 2 secondary indicator classifications: core network Speed limit status, bearer network status; the transmission link includes three secondary indicator categories: transmission link bandwidth limit status, transmission link large delay jitter status, and transmission link packet loss and disorder status; the gNB includes 3 secondary indicator categories: base station rate limit status, base station processing capability status, algorithm feature limit status; the UE and the PC connected to the UE include terminal capability status, PC performance status, TCP setting status, and software setting status;

S1.3: Set the weight of each secondary indicator classification in the node score of the corresponding link;

S1.4: Regularly score the corresponding link nodes according to the status of the secondary indicator classification; at the same time, obtain the score of each secondary indicator classification in each round of evaluation based on the score proportion weight.
A method for predicting and locating abnormality in core network and base station transmission network according to claim 1, characterized in that the specific content of step S2 is: determining the overall number of evaluations before the nth round of evaluation through historical data, and determining the number of evaluations before the nth round of evaluation. The number of abnormal times for each secondary indicator category in the overall number of evaluations is calculated by calculating the proportion of the number of abnormal times for each secondary indicator category to the total number of evaluations to determine the initial abnormal occurrence of the corresponding secondary indicator category in the nth round of evaluation. Probability a, so the initial normal occurrence probability b=1-a of the corresponding secondary indicator classification in the nth round of evaluation.
A method for predicting and locating abnormality in core network and base station transmission network according to claim 1, characterized in that the specific content of step S3 is: obtaining the classification of each secondary index in the historical data before the current nth round of evaluation. score, and perform a weighted average to obtain the average score of each secondary indicator category. Compare the score of each secondary indicator category in the current n rounds of evaluation with the average score of the corresponding secondary indicator category to determine the increase or decrease. Determine the probability c that each secondary indicator category changes from abnormal to normal in subsequent evaluations based on the interval of increase and decrease, and calculate the probability d=1-c that each secondary indicator category changes from abnormal to remain abnormal in subsequent evaluations; In the same way, determine the probability e of each secondary indicator classification from normal to abnormal in subsequent evaluations based on the interval of increase and decrease, and calculate the probability f=1 of each secondary indicator classification that changes from normal to normal in subsequent evaluations. -c.
A method for predicting and locating abnormality in core network and base station transmission network according to claim 1, characterized in that the specific content of step S4 is: predicting and calculating the abnormality of each secondary index classification in the n+1th round of evaluation. The relationship between the occurrence probability g, the normal occurrence probability h, and the initial abnormal occurrence probability a and the initial normal occurrence probability b of the nth round of the current evaluation round is: g=a*d+b*e, h=a*c+b *f.
A method for predicting and locating abnormality in core network and base station transmission network according to claim 1, further comprising:

Step S7: Select each secondary indicator prediction data and evaluation score data obtained from some rounds of evaluation in the multi-round evaluation process in step S6, generate a failure probability baseline and a score baseline, and conduct comparative analysis;

If the predicted abnormality probability increases and the score decreases, the two results are consistent and it is normal;

If the predicted abnormality probability decreases and the score increases, the two results are consistent and it is normal;

If the predicted abnormality probability increases and the score decreases, the two results do not match, which is an abnormal situation, indicating that there are hidden abnormalities in the corresponding secondary indicator classification;

If the predicted abnormality probability decreases and the score increases, the two results do not match, which is an abnormal situation, indicating that there are hidden abnormalities in the secondary indicator classification;

Check the secondary indicator classification based on the matching status of the two.