CN113411236B - Quality difference router detection method, quality difference router detection device, quality difference router detection equipment and storage medium - Google Patents

Abstract

The invention belongs to the technical field of communication, and discloses a detection method, device, equipment and storage medium of a poor quality router. The invention determines the router to be detected according to the quality poor detection instruction when receiving the quality poor detection instruction; collects the information of the attached device, the uplink state information and the information of the sending and receiving messages of the router to be detected; The status information and the information of the sent and received messages construct the feature information to be detected; based on the feature information to be detected, the router to be detected is subjected to poor quality detection through a preset poor quality detection model to obtain a poor quality detection result. Since the detection of poor quality routers is not empirical judgment, a large number of misjudgments are avoided, and the preset poor quality detection model is a pre-trained fusion model, which has high detection accuracy and does not need to be constructed in real time. It can quickly and accurately detect whether a router is Identify the poor quality router and determine the reason for the poor quality of the router, and then output the poor quality detection result.

Description

Poor quality router detection method, device, equipment and storage medium

technical field

The present invention relates to the field of communication technologies, and in particular, to a method, device, device and storage medium for detecting poor quality routers.

Background technique

Communication operators are often complained by customers because of poor signals, and a large part of them are poor signals caused by poor quality routers. For example, the routers in many homes have not been replaced for many years, the routers are old models, and the supported wireless rates are low. , the network speed is slow, resulting in user complaints; some routers have exceeded the service life standard, and the network speed is slow, resulting in user complaints; some routers have a large number of simultaneous users, and the network speed of a single user is slow, resulting in user complaints.

At present, there are two main detection methods for poor quality routers. One is to determine whether the current router is a poor quality router based on the router's time to market and network speed. The other is to use data mining to classify the detected data. A large amount of data is used to construct a feature library of poor-quality routers, and the feature library is used to determine whether a router is a poor-quality router.

The shortcomings of the first detection method of poor quality routers are very obvious: the standard is too empirical, and a large number of routers whose speed still meets the contract speed and are stable in operation may be misjudged as poor quality routers; for different models and different scenarios, it is difficult to analyze Poor quality router network speed; it is impossible to determine the specific cause of the poor router quality.

The second detection method of poor-quality routers needs to build a signature database in advance, needs to collect a large amount of data, and consumes a lot of server resources to build a signature database, which is extremely important for the completeness of data sources and the sufficiency of server resources for the provincial branches of communication operators. High requirements are difficult and costly to implement, and since a large amount of server resources are required to build a feature library, the overall process takes a long time, and it is difficult to meet the needs of real-time judgment.

The above content is only used to assist the understanding of the technical solutions of the present invention, and does not mean that the above content is the prior art.

SUMMARY OF THE INVENTION

The main purpose of the present invention is to provide a detection method, device, equipment and storage medium for poor quality routers, aiming to solve the technical problem that the prior art cannot detect poor quality routers quickly and accurately.

To achieve the above object, the invention provides a kind of poor quality router detection method, described method comprises the following steps:

When receiving the poor quality detection instruction, determine the router to be detected according to the poor quality detection instruction;

Collect the information of the connected devices, the uplink status information and the information of the sent and received packets of the router to be detected;

constructing feature information to be detected according to the downlink device information, the uplink status information, and the sent and received message information;

Based on the feature information to be detected, quality poor detection is performed on the router to be detected by using a preset quality poor detection model to obtain a poor quality detection result.

Optionally, before the step of determining the router to be detected according to the quality difference detection instruction when the quality difference detection instruction is received, the method further includes:

acquiring data samples stored in the preset sample library, and constructing a data sample set according to the data samples;

Perform multiple training on the first type of initial model in the initial model set according to the data sample set to obtain multiple target decision tree models;

Perform training on other types of initial models in the initial model set according to the data sample set to obtain multiple models to be fused;

Model fusion is performed on the multiple target decision tree models and the multiple models to be fused to obtain a preset quality poor detection model.

Optionally, before the step of acquiring the data samples stored in the preset sample library and constructing a data sample set according to the data samples, the method further includes:

Periodically collect the router operation information of users of each platform through the intelligent networking platform;

A data sample is constructed according to the router operation information, and the data sample is stored in a preset sample library.

Optionally, before the step of constructing a data sample according to the router operation information and storing the data sample in a preset sample library, the method further includes:

Detecting whether the router operation information is missing data;

When the router operation information is missing data, obtain historical operation information corresponding to the router operation information;

Perform data completion on the router operation information according to the historical operation information to obtain the completed router operation information;

Correspondingly, the step of constructing a data sample according to the router operation information and storing the data sample in a preset sample library includes:

A data sample is constructed according to the completed router operation information, and the data sample is stored in a preset sample library.

Optionally, before the step of constructing a data sample according to the router operation information and storing the data sample in a preset sample library, the method further includes:

performing data reduction on the router operation information to obtain reduced router information;

Correspondingly, the step of constructing a data sample according to the router operation information and storing the data sample in a preset sample library includes:

A data sample is constructed according to the reduced router information, and the data sample is stored in a preset sample library.

Optionally, the step of performing multiple training on the first type of initial model in the initial model set according to the data sample set to obtain multiple target decision tree models includes:

constructing a plurality of data sample subsets according to the data sample set;

Based on the first preset number of random seeds and the second preset number of feature subsets, the first type of initial model in the initial model set is trained multiple times according to each data sample subset to obtain multiple target decision tree models.

Optionally, before the step of performing model fusion of the multiple target decision tree models and the multiple models to be fused to obtain a preset quality difference detection model, the method further includes:

Based on the maximum model evaluation score, determine the model fusion weights corresponding to each target decision tree model and each model to be fused;

Correspondingly, the step of performing model fusion of the multiple target decision tree models and the multiple models to be fused to obtain a preset quality poor detection model includes:

Based on the model fusion weight, model fusion is performed on the multiple target decision tree models and the multiple models to be fused to obtain a preset quality difference detection model.

In addition, in order to achieve the above purpose, the present invention also provides a poor quality router detection device, the poor quality router detection device includes the following modules:

an instruction response module, configured to determine the router to be detected according to the quality difference detection instruction when receiving the quality difference detection instruction;

an information acquisition module, which is used to collect the information of the attached device, the uplink status information and the information of the sent and received packets of the router to be detected;

an information construction module, configured to construct feature information to be detected according to the down-connected device information, the uplink status information and the sent and received message information;

The quality difference detection module is configured to perform quality difference detection on the router to be detected by using a preset quality difference detection model based on the feature information to be detected, so as to obtain a quality difference detection result.

In addition, in order to achieve the above object, the present invention also provides a poor-quality router detection device, the poor-quality router detection device includes: a processor, a memory, and a quality device stored in the memory and running on the processor. A poor router detection program, when the poor router detection program is executed by the processor, implements the steps of the above-described poor router detection method.

In addition, in order to achieve the above object, the present invention also provides a computer-readable storage medium, where a poor-quality router detection program is stored on the computer-readable storage medium, and when the poor-quality router detection program is executed, the above-mentioned quality The steps of the poor router detection method.

The invention determines the router to be detected according to the quality poor detection instruction when receiving the quality poor detection instruction; collects the information of the attached device, the uplink state information and the information of the sending and receiving messages of the router to be detected; The status information and the information of the sent and received messages construct the feature information to be detected; based on the feature information to be detected, the router to be detected is subjected to poor quality detection through a preset poor quality detection model to obtain a poor quality detection result. Since the detection of poor quality routers is not empirical judgment, a large number of misjudgments are avoided, and the preset poor quality detection model is a pre-trained fusion model, which has high detection accuracy and does not need to be constructed in real time. It can quickly and accurately detect whether a router is It is a poor router and determines the reason for the poor quality of the router, and then outputs the poor quality detection result. The user or the service personnel of the communication operator can determine whether the router to be tested is a poor quality router according to the poor quality detection result, and if the router to be detected is a poor quality router The reasons for the poor quality are clearly identified at the same time, so as to facilitate the follow-up processing.

Description of drawings

1 is a schematic structural diagram of an electronic device of a hardware operating environment involved in an embodiment of the present invention;

2 is a schematic flowchart of a first embodiment of a poor quality router detection method according to the present invention;

3 is a schematic flowchart of a second embodiment of a poor quality router detection method according to the present invention;

4 is a schematic diagram of a model parameter tuning process according to an embodiment of the poor quality router detection method of the present invention;

FIG. 5 is a schematic diagram of model fusion according to an embodiment of a method for detecting only poor routers according to the present invention;

FIG. 6 is a structural block diagram of a first embodiment of a poor quality router detection apparatus according to the present invention.

The realization, functional characteristics and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments.

Detailed ways

It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

Referring to FIG. 1 , FIG. 1 is a schematic structural diagram of a poor quality router detection device of a hardware operating environment involved in an embodiment of the present invention.

As shown in FIG. 1 , the electronic device may include: a processor 1001 , such as a central processing unit (Central Processing Unit, CPU), a communication bus 1002 , a user interface 1003 , a network interface 1004 , and a memory 1005 . Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. Optionally, the network interface 1004 may include a standard wired interface and a wireless interface (eg, a Wireless-Fidelity (WI-FI) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM) memory, or may be a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001 .

Those skilled in the art can understand that the structure shown in FIG. 1 does not constitute a limitation on the electronic device, and may include more or less components than the one shown, or combine some components, or arrange different components.

As shown in FIG. 1 , the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and a poor-quality router detection program.

In the electronic device shown in FIG. 1, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the electronic device of the present invention can be set in In the poor quality router detection device, the electronic device invokes the poor quality router detection program stored in the memory 1005 through the processor 1001, and executes the poor quality router detection method provided by the embodiment of the present invention.

An embodiment of the present invention provides a method for detecting poor quality routers. Referring to FIG. 2 , FIG. 2 is a schematic flowchart of a first embodiment of a method for detecting poor quality routers according to the present invention.

In this embodiment, the method for detecting poor quality routers includes the following steps:

Step S10: When receiving the quality difference detection instruction, determine the router to be detected according to the quality difference detection instruction.

It should be noted that the execution subject of this embodiment may be the poor-quality router detection device, and the poor-quality router detection device may be an electronic device such as a computer or a server, or may be other devices with the same or similar functions. The example is not limited to this. In this embodiment and the following embodiments, a poor-quality router detection device is used as an example to describe the poor-quality router detection method of the present invention.

It should be noted that the poor-quality detection instruction may be an instruction sent by other devices to the poor-quality router detection device, and the poor-quality router detection instruction may carry a router identifier. According to the poor-quality detection instruction, it is determined that the router to be detected may be an extracted quality router. The router identifier in the difference detection instruction is used to find the router to be detected according to the router identifier.

Step S20: Collect the information of the attached device, the uplink state information and the information of the sent and received packets of the router to be detected.

It should be noted that the hanging device may be a device connected to the router to be detected, such as an electronic device such as a mobile phone, a computer, and a tablet computer. The information of the attached device can include: the MAC address of the attached device, the connection duration, the value of the access frequency band (2.4G or 5G), the signal strength, the current negotiated reception rate and transmission rate, and the real-time uplink reception rate and transmission rate. and other information, the use of the router to be detected, the frequency band preference and the WI-FI quality information of the connected device can be reflected through the information of the connected device. The uplink status information can include: the management MAC address used in the uplink, the frequency band (2.4G or 5G) used for the WLAN uplink, the uplink signal strength during the uplink, the negotiated reception rate and transmission rate of the uplink, and the real-time uplink Information such as the receiving rate and the sending rate can reflect the running status of the gateway device connected to the router through the uplink status information, which can be used to assist in determining the reason for the poor quality of the router. The information of sending and receiving messages can include: the total number of bytes sent and received, the total number of packets sent and received, the number of error packets sent and the number of error packets received, the number of packets discarded during transmission and reception, and other information. By sending and receiving message information, it can comprehensively reflect the historical operation of the router, which can be used as a supplement to real-time data such as information on connected devices.

Step S30: Constructing feature information to be detected according to the information of the connected device, the information of the uplink state, and the information of the sent and received packets.

It can be understood that the feature information to be detected is the data used to input the preset quality difference detection model for analysis, and the preset quality difference detection model may have format requirements for the input data. The feature information to be detected is constructed based on the connection status information and the message information sent and received, so that the quality difference detection can be carried out through the preset quality difference detection model.

In actual use, constructing the feature information to be detected according to the information of the connected device, the information of the uplink state and the information of the sent and received messages may be to extract the features of the information of the connected device, the information of the connected state and the information of the sent and received messages, according to the extracted features. The information constructs the feature information to be detected.

Step S40: Based on the feature information to be detected, perform a quality difference detection on the router to be detected by using a preset quality difference detection model to obtain a quality difference detection result.

It should be noted that the preset quality difference detection model may be obtained by fusing multiple basic models, wherein the multiple basic models may be obtained by training multiple different algorithm models, and the preset quality difference detection model may be managed by Personnel pre-set. The preset poor quality detection model may determine whether the router to be detected is a poor quality router according to the input feature information to be detected, and output a corresponding poor quality detection result. The poor quality detection result can include the judgment result of the poor quality router and the reason for the poor quality of the router, wherein the judgment result of the poor quality router can be represented by 0 or 1, where 0 indicates that the router is not of poor quality, 1 indicates that the router is judged to be of poor quality, and the router is of poor quality. The reasons for poor quality can include WiFi coverage problems (weak WiFi signal of the connected device), WiFi interference problems (excessive 2.4G/5G co-channel interference), WiFi configuration problems (unreasonable gateway/router parameter configuration), WiFi usage problems (below The number of hanging devices is too large, the bandwidth usage is saturated), etc.

It is understandable that the poor-quality router detection device can also display the poor-quality detection results after determining the poor-quality detection results, so that users or service personnel of communication operators can determine whether the routers to be detected are poor-quality routers, and the routers to be detected are of poor quality. When the router is a poor quality router, clarify the reason for the poor quality, so as to facilitate subsequent processing.

In this embodiment, when a quality-bad detection instruction is received, the router to be detected is determined according to the quality-bad detection instruction; Based on the feature information to be detected, the router to be detected is subjected to poor quality detection through a preset poor quality detection model to obtain a poor quality detection result. Since the detection of poor quality routers is not empirical judgment, a large number of misjudgments are avoided, and the preset poor quality detection model is a pre-trained fusion model, which has high detection accuracy and does not need to be constructed in real time. It can quickly and accurately detect whether a router is It is a poor router and determines the reason for the poor quality of the router, and then outputs the poor quality detection result. The user or the service personnel of the communication operator can determine whether the router to be tested is a poor quality router according to the poor quality detection result, and if the router to be detected is a poor quality router The reasons for the poor quality are clearly identified at the same time, so as to facilitate the follow-up processing.

Referring to FIG. 3 , FIG. 3 is a schematic flowchart of a second embodiment of a poor quality router detection method according to the present invention.

Based on the above-mentioned first embodiment, before the step S10, the method for detecting a poor-quality router in this embodiment further includes:

Step S01: Acquire data samples stored in a preset sample library, and construct a data sample set according to the data samples.

It should be noted that the preset sample library may be a preset database for storing data samples, and a large number of data samples constructed according to pre-collected router operation information may be stored therein. The data samples may include router operation information and manually marked information labels, and the information labels may include whether the router is of poor quality and the reasons for the poor quality of the router. Acquiring data samples stored in a preset sample library and constructing a data sample set based on the data samples may be acquiring a preset number of data samples in the preset sample library, and constructing a data sample set according to a combination of the acquired data samples. The preset number of samples can be set by the administrator according to actual needs, for example, the preset number of samples is set to 10,000.

In actual use, considering the timeliness of data, data samples newly added to the preset sample library may be used as much as possible when acquiring data samples from the preset sample library.

Further, in order to reduce labor costs, before step S01 in this embodiment, it may further include:

The router operation information of each platform user is periodically collected through the intelligent networking platform; data samples are constructed according to the router operation information, and the data samples are stored in a preset sample library.

It should be noted that the router operation information may include information such as the information of the devices attached to the router, the connection status information, and the information of sending and receiving messages. The intelligent networking platform may be the network management platform set by the communication operator. The routers of users of each platform are Connected to the intelligent networking platform, the intelligent networking platform can maintain a heartbeat connection with the router through the agreed protocol, and can establish a connection with the router of the platform user through polling tasks and reverse triggering, and periodically collect the router operation information of each platform router. , since there is no need to manually collect router operation information, labor costs can be reduced to a certain extent.

It can be understood that after obtaining the router operation information, the administrator may not have time to further process the collected router operation information. Therefore, the router operation information can also be saved in a preset database, and can be retrieved from the database later. Read router operation information for further processing.

It should be noted that, constructing data samples according to the router operation information can be performed by manually analyzing the router operation information to determine whether it is a poor quality router, determine the reason for the poor quality, and determine the information label, and then perform feature extraction on the router operation information. Obtain feature information, and construct data samples according to feature information and information labels.

Further, due to the influence of various factors in the process of data collection, some data may be missing in the collected router operation information. The accuracy and generalization ability of the trained model, but if the router operation information of the missing data is directly deleted, the data may be incomplete. Before the step of storing the data samples in the preset sample library, it may also include:

Detecting whether the router operation information is missing data; when the router operation information is missing data, obtain historical operation information corresponding to the router operation information; perform data completion on the router operation information according to the historical operation information, to Obtain the completed router operation information;

Correspondingly, the step of constructing a data sample according to the router operation information and storing the data sample in a preset sample library may include:

A data sample is constructed according to the completed router operation information, and the data sample is stored in a preset sample library.

It should be noted that, to detect whether the router operation information is real data, the data fields in the router operation information can be compared with the preset data fields to be collected, and whether there is a field missing, if there is a missing field, the missing data can be determined. If there is no missing field, you can also obtain the value corresponding to each field, and determine whether the value corresponding to each field exists. data.

It should be noted that when completing data, there are generally two ways. The first is to use an algorithm to predict the value of the missing part of the data, and fill in the missing part of the data according to the predicted value, for example: through the decision tree algorithm Or naive Bayes and other algorithms to predict the data, and supplement the missing values according to the predicted values; the second is to use the average value to replace the missing values, for example: obtain the data collected before and after the data. Collected data, calculate the average value of multiple data, and supplement according to the average value.

In actual use, the historical operation information corresponding to the router operation information can be obtained, the data of the missing part of the currently collected router operation information can be inferred according to the historical operation information, and the missing data can be supplemented according to the inferred data.

It should be noted that, because the data involved in this embodiment is mainly router operation information, and data such as the rate of the router generally does not change too much abnormally in a short period of time, that is, the amount of change is small, and the missing data of the attached device is generally The proportion is not high, therefore, the average value can also be used to replace the missing value, then the router operation information of the missing data can be obtained. The mean value of , and the missing data are supplemented according to the mean value.

Further, in order to improve the data processing efficiency, before the steps of constructing data samples according to the router operation information and storing the data samples in the preset sample library described in this embodiment, the method may further include:

performing data reduction on the router operation information to obtain reduced router information;

Correspondingly, the step of constructing a data sample according to the router operation information and storing the data sample in a preset sample library includes:

A data sample is constructed according to the reduced router information, and the data sample is stored in a preset sample library.

It should be noted that in the scenario of a huge amount of data (above one million), it takes a long time to analyze and mine the data, which consumes a lot of server performance and resources. Therefore, the collected data can be compressed to simplify the data. Data reduction technology does not affect data integrity, but can help improve data quality and data processing efficiency.

In this embodiment, data reduction may include at least one of the following four ways:

(1) Data dimensionality reduction, delete part of the data in the router operation information that is very low or irrelevant to the router quality difference analysis, for example: delete the MAC address of the attached device in the attached device information.

(2) Data compression, the data size of the collected router operation information can be reduced by using the coding mechanism.

(3) Data reduction, using smaller data to replace some larger data.

(4) Generation of concept level, use higher-level attributes or concepts to combine less powerful data, for example: integrate the receiving and sending rates of multiple connected devices into one data, instead of the corresponding data of each connected device A receive and transmit rate.

Step S02: Perform multiple training on the first type of initial model in the initial model set according to the data sample set to obtain multiple target decision tree models.

It should be noted that the initial model set may store multiple initial models of different types, and the different types of initial models may be models constructed according to different algorithms. Among them, the first type of initial model may be a model constructed according to a gradient boosting decision tree (Gradient Boosting Decision Tree, GBDT) algorithm, wherein the gradient boosting decision tree algorithm is an iterative decision tree algorithm, and the core of the algorithm is to pass multiple weak The learner is repeatedly trained and iterated, and finally multiple weak learners are combined into a strong learner, and the performance of the strong learner will be greatly improved compared to a single weak learner.

Further, in order to improve the effect of model fusion, step S02 in this embodiment may include:

Build a plurality of data sample subsets according to the data sample set; based on the first preset number of random seeds and the second preset number of feature subsets, according to each data sample subset, the initial model set of the first type of initial model Do multiple trainings to obtain multiple target decision tree models.

It should be noted that, constructing multiple data sample subsets according to the data sample set may be to select data samples with a preset number of subset samples from the data sample set to construct a data sample subset, which can be executed repeatedly. The difference of the decision tree model can improve the effect of model fusion, and try to ensure that the data samples in each data sample subset are different.

It should be noted that the random seed will affect the samples examined by a single tree in the decision tree and the features selected when splitting nodes. Fixing the random seed during the training process can ensure that the results are the same when the random seed remains unchanged. The subset can be set manually by grouping the features to indicate which part of the features in the data sample is used for model training. The first preset number and the second preset number can be set by the administrator according to actual needs, for example, the first preset number is set to 30, and the second preset number is set to 10.

In actual use, in order to ensure that each target decision tree model obtained by training has a certain difference as much as possible, so as to obtain the best training results, each random seed can be different from each other, and each feature subset is also different from each other.

In actual use, the first type of initial model can be a model constructed according to the GBDT algorithm. The parameters of the GBDT algorithm are shown in Table 1 below:

Table 1GBDT parameters

In the actual training process, the parameter tuning process of the GBDT model can be shown in Figure 4.

In actual use, a data sample subset can be selected from multiple data sample subsets, and then based on the selected data sample subset, the first preset number of random seeds and the second preset number of feature subsets One type of initial model is trained, multiple base learners are obtained, and then multiple base learners are fused in equal proportions (because the algorithms used are consistent and the accuracy is basically similar, so they can be fused in proportion according to the voting method, without setting weights) to obtain a target decision tree model. By repeating the above steps, multiple target decision tree models can be obtained according to multiple data sample subsets.

For example: Assuming that the data sample subsets are A, B, and C, the random seeds are S1, S2, S3, and S4, and the feature subsets are T1, T2, and T3, the A data sample can be selected first. Subset, the first type of initial model is trained according to A and S1, and the parameters are continuously adjusted to obtain the basic learner A-S1. After multiple executions, A-S1, A-S2, A-S3, A- S4, A-T1, A-T2, A-T3 have a total of 7 basic learners, and then fuse the 7 basic learners in equal proportions to obtain the target decision tree model MA corresponding to the subset of data samples, and repeat the above steps , and then obtain MB and MC to obtain multiple target decision tree models.

Step S03: Train other types of initial models in the initial model set according to the data sample set to obtain a plurality of models to be fused.

It should be noted that there can be multiple other types of initial models, all of which can be initial models constructed by different algorithms. For example, there can be 3 other types of initial models, which are AdaBoost algorithm, vector machine algorithm or random forest algorithm. (Random Forest, RF) to build the initial model.

Step S04: Perform model fusion on the multiple target decision tree models and the multiple models to be fused to obtain a preset quality difference detection model.

It should be noted that the voting method (Voting) can be used to fuse multiple target decision tree models and multiple models to be fused. The voting method (Soft Voting) is not limited in this embodiment.

It can be understood that by performing model fusion of multiple target decision tree models and multiple models to be fused, the obtained preset poor quality detection model has better comprehensive performance than a single target decision tree model or a single model to be fused.

Further, since each model to be fused is obtained from initial model training based on different algorithms, and it is different from the algorithm used by the target decision tree model, the accuracy rate will be different. The effect of the fused model is not optimal. In order to ensure the use effect of the fused model, the multiple target decision tree models and the multiple models to be fused are model-fused to obtain a preset quality difference detection. Before the steps of the model, you can also include:

Based on the maximum model evaluation score, determine the model fusion weights corresponding to each target decision tree model and each model to be fused;

Correspondingly, the step of performing model fusion of the multiple target decision tree models and the multiple models to be fused to obtain a preset quality poor detection model includes:

Based on the model fusion weight, model fusion is performed on the multiple target decision tree models and the multiple models to be fused to obtain a preset quality difference detection model.

It should be noted that the model evaluation score can use the F1 score (F1 Score), and the F1 score can take into account both the precision rate and the recall rate of the model. Based on the maximum model evaluation score, determine the model fusion weights corresponding to each target decision tree model and each model to be fused. First, set the weight of each model to 1, then perform fusion, and use the test set for testing and calculation. F1 score, and then adjust the weight of each model according to the preset step size, continuously test and calculate the F1 score, and use the weight value of each model when the F1 score is the largest as the model fusion weight of each model.

It can be understood that, based on the model fusion weight, the model fusion of multiple target decision tree models and multiple models to be fused, the obtained preset quality difference detection model can ensure that its F1 score is the largest, that is, the comprehensive performance of the model is the best. , use the best effect.

For ease of understanding, reference is made to Fig. 5 for description, but this scheme is not limited. Fig. 5 is a schematic diagram of model fusion, in which 30 different random seeds and 10 different feature subsets are set first, and according to the data sample subsets According to the initial model constructed by the GBDT algorithm, the first target decision tree model is obtained, and it is used as the voting classifier (GBDT1), and then different subsets of data samples are used to train to obtain the voting classifier (GBDT2) and Voting classifier (GBDT3), and then train the initial model constructed according to the AdaBoost algorithm according to the data sample set, obtain the model to be fused, and use it as the voting classifier (Ada), and then use the data sample set to construct according to the random forest algorithm. The initial model is trained, the model to be fused is obtained, and it is used as a voting classifier (RF), and then the model fusion weight is determined, and each model is fused according to the model fusion weight, and the fused voting classifier is obtained, which is used as the final result ( FINAL RESULT) to obtain a preset quality poor detection model.

In this embodiment, the data samples stored in the preset sample library are acquired, and a data sample set is constructed according to the data samples; and the first type of initial model in the initial model set is trained multiple times according to the data sample set to obtain multiple target decision tree model; according to the data sample set, other types of initial models in the initial model set are trained to obtain multiple models to be fused; the multiple target decision tree models and the multiple models to be fused are performed Model fusion to obtain a preset poor quality detection model. Because it is not a model constructed by a single algorithm, but a fusion of models constructed by a variety of different algorithms after training, the comprehensive performance of the preset poor quality detection model is better.

In addition, an embodiment of the present invention also provides a storage medium, where a poor-quality router detection program is stored on the storage medium, and when the poor-quality router detection program is executed by a processor, the method for detecting a poor-quality router as described above is implemented. step.

Referring to FIG. 6 , FIG. 6 is a structural block diagram of a first embodiment of a poor quality router detection apparatus according to the present invention.

As shown in FIG. 6 , the device for detecting poor quality routers provided by the embodiment of the present invention includes:

The instruction response module 601 is configured to determine the router to be detected according to the quality difference detection instruction when receiving the quality difference detection instruction;

an information acquisition module 602, configured to collect the information of the connected devices, the uplink status information and the information of the sent and received messages of the router to be detected;

an information construction module 603, configured to construct feature information to be detected according to the connected device information, the uplink status information and the sent and received message information;

The quality difference detection module 604 is configured to perform quality difference detection on the router to be detected by using a preset quality difference detection model based on the feature information to be detected, so as to obtain a quality difference detection result.

In this embodiment, when a quality-bad detection instruction is received, the router to be detected is determined according to the quality-bad detection instruction; Based on the feature information to be detected, the router to be detected is subjected to poor quality detection through a preset poor quality detection model to obtain a poor quality detection result. Since the detection of poor quality routers is not empirical judgment, a large number of misjudgments are avoided, and the preset poor quality detection model is a pre-trained fusion model, which has high detection accuracy and does not need to be constructed in real time. It can quickly and accurately detect whether a router is It is a poor router and determines the reason for the poor quality of the router, and then outputs the poor quality detection result. The user or the service personnel of the communication operator can determine whether the router to be detected is a poor quality router according to the poor quality detection result, and the router to be detected is a poor quality router. The reasons for the poor quality are clearly identified at the same time, so as to facilitate the follow-up processing.

Further, the instruction response module 601 is further configured to obtain the data samples stored in the preset sample library, and construct a data sample set according to the data samples; according to the data sample set, the initial model set of the first type in the initial model set is Perform multiple trainings to obtain multiple target decision tree models; train other types of initial models in the initial model set according to the data sample set to obtain multiple models to be fused; combine the multiple target decision tree models and the multiple models to be fused to perform model fusion to obtain a preset quality difference detection model.

Further, the instruction response module 601 is further configured to periodically collect the router operation information of each platform user through the intelligent networking platform; construct a data sample according to the router operation information, and store the data sample into a preset sample in the library.

Further, the instruction response module 601 is further configured to detect whether the router operation information is missing data; when the router operation information lacks data, obtain historical operation information corresponding to the router operation information; according to the historical operation information The information performs data completion on the router operation information to obtain the completed router operation information;

The instruction response module 601 is further configured to construct a data sample according to the completed router operation information, and store the data sample in a preset sample library.

Further, the instruction response module 601 is further configured to perform data reduction on the router operation information to obtain reduced router information;

The instruction response module 601 is further configured to construct a data sample according to the reduced router information, and store the data sample in a preset sample library.

Further, the instruction response module 601 is further configured to construct a plurality of data sample subsets according to the data sample set; based on the first preset number of random seeds and the second preset number of feature subsets, according to each data The sample subsets respectively perform multiple training on the first type of initial models in the initial model set to obtain multiple target decision tree models.

Further, the instruction response module 601 is further configured to determine the model fusion weight corresponding to each target decision tree model and each to-be-fused model based on the maximum model evaluation score;

The instruction response module 601 is further configured to perform model fusion of the multiple target decision tree models and the multiple models to be fused based on the model fusion weight to obtain a preset quality difference detection model.

It should be understood that the above are only examples, and do not constitute any limitation to the technical solutions of the present invention. In specific applications, those skilled in the art can make settings as required, which is not limited by the present invention.

It should be noted that the above-described workflow is only illustrative, and does not limit the protection scope of the present invention. In practical applications, those skilled in the art can select some or all of them to implement according to actual needs. The purpose of the solution in this embodiment is not limited here.

In addition, for technical details that are not described in detail in this embodiment, reference may be made to the detection method for a poor-quality router provided by any embodiment of the present invention, and details are not repeated here.

Furthermore, it should be noted that, herein, the terms "comprising", "comprising" or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, method, article or system comprising a series of elements includes not only those elements, but also other elements not expressly listed or inherent to such a process, method, article or system. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article or system that includes the element.

The above-mentioned serial numbers of the embodiments of the present invention are only for description, and do not represent the advantages or disadvantages of the embodiments.

From the description of the above embodiments, those skilled in the art can clearly understand that the method of the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is better implementation. Based on such understanding, the technical solutions of the present invention can be embodied in the form of software products in essence or the parts that make contributions to the prior art, and the computer software products are stored in a storage medium (such as a read-only memory). , ROM)/RAM, magnetic disk, optical disk), including several instructions to make a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) execute the methods described in the various embodiments of the present invention.

The above are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied in other related technical fields , are similarly included in the scope of patent protection of the present invention.

Claims (9)

1. A quality difference router detection method is characterized by comprising the following steps:

when a quality difference detection instruction is received, determining a router to be detected according to the quality difference detection instruction;

acquiring information of the lower hanging device, information of the upper connection state and information of a message to be received of the router to be detected;

constructing characteristic information to be detected according to the information of the lower hanging device, the information of the upper connection state and the information of the receiving and sending messages;

based on the characteristic information to be detected, performing quality difference detection on the router to be detected through a preset quality difference detection model to obtain a quality difference detection result;

before the step of determining the router to be detected according to the quality difference detection instruction when the quality difference detection instruction is received, the method further includes:

acquiring data samples stored in a preset sample library, and constructing a data sample set according to the data samples;

training the first type initial model in the initial model set for multiple times according to the data sample set to obtain a plurality of target decision tree models;

training other types of initial models in the initial model set according to the data sample set to obtain a plurality of models to be fused;

and carrying out model fusion on the target decision tree models and the models to be fused so as to obtain a preset quality difference detection model.

2. The method for detecting a poor quality router according to claim 1, wherein before the step of obtaining the data samples stored in the preset sample library and constructing the data sample set according to the data samples, the method further comprises:

periodically acquiring router operation information of each platform user through an intelligent networking platform;

and constructing a data sample according to the router operation information, and storing the data sample into a preset sample library.

3. The method for detecting a poor quality router according to claim 2, wherein before the step of constructing the data sample according to the router operation information and storing the data sample in a preset sample library, the method further comprises:

detecting whether the router operation information is lack of data;

when the router running information lacks data, acquiring historical running information corresponding to the router running information;

performing data completion on the router operation information according to the historical operation information to obtain completed router operation information;

correspondingly, the step of constructing a data sample according to the router operation information and storing the data sample in a preset sample library includes:

and constructing a data sample according to the complemented router operation information, and storing the data sample into a preset sample library.

4. The method for detecting a poor quality router according to claim 2, wherein before the step of constructing the data sample according to the router operation information and storing the data sample in a preset sample library, the method further comprises:

performing data reduction on the router operation information to obtain reduced router information;

correspondingly, the step of constructing a data sample according to the router operation information and storing the data sample in a preset sample library includes:

and constructing a data sample according to the reduced router information, and storing the data sample into a preset sample library.

5. The method of detecting a poor quality router of claim 1 wherein the step of training a first type of initial model in an initial set of models a plurality of times based on the set of data samples to obtain a plurality of objective decision tree models comprises:

constructing a plurality of data sample subsets from the set of data samples;

and training the first type initial model in the initial model set for multiple times respectively according to the data sample subsets based on the random seeds with the first preset number and the feature subsets with the second preset number to obtain a plurality of target decision tree models.

6. The method according to claim 1, wherein before the step of performing model fusion on the target decision tree models and the models to be fused to obtain the preset quality difference detection model, the method further comprises:

determining model fusion weights corresponding to each target decision tree model and each model to be fused according to the maximum model evaluation value;

correspondingly, the step of performing model fusion on the target decision tree models and the models to be fused to obtain a preset quality difference detection model includes:

and performing model fusion on the target decision tree models and the models to be fused based on the model fusion weight to obtain a preset quality difference detection model.

7. A quality difference router detection apparatus, comprising:

the instruction response module is used for determining the router to be detected according to the quality difference detection instruction when the quality difference detection instruction is received;

the information acquisition module is used for acquiring information of the lower hanging device, information of the upper connection state and information of a receiving and sending message of the router to be detected;

the information construction module is used for constructing the characteristic information to be detected according to the information of the lower hanging device, the information of the upper connection state and the information of the receiving and sending messages;

the quality difference detection module is used for carrying out quality difference detection on the router to be detected through a preset quality difference detection model based on the characteristic information to be detected so as to obtain a quality difference detection result;

the instruction response module is further used for obtaining data samples stored in a preset sample library and constructing a data sample set according to the data samples; training the first type initial model in the initial model set for multiple times according to the data sample set to obtain a plurality of target decision tree models; training other types of initial models in the initial model set according to the data sample set to obtain a plurality of models to be fused; and carrying out model fusion on the target decision tree models and the models to be fused so as to obtain a preset quality difference detection model.

8. A quality-difference router detection apparatus, characterized in that the quality-difference router detection apparatus comprises: processor, memory and a quality router detection program stored on the memory and executable on the processor, the quality router detection program when executed by the processor implementing the steps of the quality router detection method according to any of claims 1-6.

9. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a quality-difference router detection program, which when executed performs the steps of the quality-difference router detection method according to any one of claims 1-6.

Images