CN109787846A - A kind of 5G network service quality exception monitoring and prediction technique and system - Google Patents

Abstract

The present invention proposes a kind of 5G network service quality exception monitoring and prediction technique and system, belongs to technical field of communication network.The system comprises: data acquisition module: for acquiring 5G network service quality data and network KPI performance monitoring data；Data processing module: for the network service quality data to be pre-processed and are marked；Qos data memory module: for storing the network service quality data after the label；Model training module: establishing Supervised machine learning model, and training obtains QoS exception monitoring device and QoS predicting abnormality device；QoS exception monitoring device: current 5G network service quality data are monitored；QoS predicting abnormality device: prediction future 5G network service quality data exception；Qos policy decision-making module: for marking and storing abnormal data, abnormal results are reported.The present invention can be the quality of service guarantee of the 5G network user, improve service quality.

Description

A 5G network service quality abnormal monitoring and prediction method and system

technical field

The invention belongs to the technical field of communication networks, and in particular relates to a method and system for abnormal monitoring and prediction of 5G network service quality based on a decision tree.

Background technique

Network Quality of Service (QoS, Quality of Service) is a necessary support to ensure network performance. Traditional network service quality assurance adopts the differentiated service Diffserv model or the integrated service Interserv model. Diffserv cannot guarantee global optimization, and the Interserv model involves complex signaling control, so the existing network can only provide best-effort QoS without guaranteeing QoS.

Under the scenarios of massive device connections, ultra-high traffic density, ultra-high connection density and ultra-high mobility, there are huge challenges in how 5G networks can meet user services. Traditional network QoS architectures cannot adapt to complex and dynamic 5G network application scenarios.

SUMMARY OF THE INVENTION

In view of this, the present invention proposes a 5G network service quality abnormality monitoring and prediction method and system, which are used to monitor network service abnormality and improve network service quality.

In a first aspect of the present invention, a method for monitoring and predicting abnormal service quality of a 5G network is proposed, and the method includes:

S1. Collect 5G network service quality data and network KPI performance monitoring data, where the network service quality data includes user terminal data, access network data, and core network data;

S2, preprocessing and marking the network service quality data;

S3, storing the marked network service quality data in a QoS database;

S4. Use the marked network service quality data in the QoS database as a data set, construct a supervised machine learning model, and use the data set to train the supervised machine learning model to obtain a QoS abnormality monitor and QoS anomaly predictor;

S5, using the QoS abnormal monitor to monitor the current 5G network service quality data in real time, and send the monitored abnormal data to the QoS policy decision module;

S6. Use the QoS anomaly predictor to predict future 5G network service quality data anomalies, and send the predicted anomalous data to the QoS policy decision module;

S7. The QoS policy decision module marks and stores the abnormal data, updates the QoS database, reports the abnormal result, and makes a decision based on the abnormal data to drive the execution of the decision.

Optionally, in step S1: the user terminal data includes: hardware data of the user terminal, software model version, installed applications, terminal position, moving direction, speed, consumed CPU, memory resources, and alarm logs; The network access data includes: base station distribution, antenna channel mode, spectrum usage, physical resource and virtual resource usage, air interface signaling, and alarm logs; the core network data includes: user service quality agreement, network slice resource usage, core network signaling , an alarm log; the network KPI performance monitoring data includes: network bandwidth, delay, and jitter.

Optionally, the specific process of step S2 is: cleaning and unifying the format of the collected network service quality data, and marking the network service quality data in combination with the network KPI performance monitoring data, where the marking includes: normal and abnormal.

Optionally, in the step S4, the supervised machine learning model adopts a decision tree algorithm.

Optionally, in step S6, the QoS anomaly predictor predicts future 5G network service quality data anomalies according to current 5G network service quality data and historical 5G network service quality data, and the historical 5G network service quality data is the Historical network quality of service data records saved in the QoS database.

Optionally, in the step S7, the QoS policy decision module marking and storing the abnormal data is specifically: automatically marking the results of the QoS abnormality monitor and the QoS abnormality predictor, and marking the new The data is stored in the QoS database, and the network service quality data in the QoS database is updated.

A second aspect of the present invention provides a 5G network service quality abnormal monitoring and prediction system, the system comprising:

Data collection module: used to collect 5G network service quality data and network KPI performance monitoring data, the network service quality data includes user terminal QoS data, access network QoS data, and core network QoS data;

Data processing module: used to preprocess and mark the network service quality data;

QoS data storage module: used to store the marked network service quality data;

Model training module: used to use the marked network service quality data in the QoS database as a data set of a supervised machine learning model, build a supervised machine learning model, train the supervised machine learning model, and obtain QoS Anomaly Monitor and QoS Anomaly Predictor

QoS abnormal monitor: used to monitor the current 5G network service quality data in real time, and send the monitored abnormal data to the QoS policy decision module;

QoS anomaly predictor: used to predict future 5G network service quality data anomalies according to the current 5G network service quality data and the historical 5G network service quality data in the QoS data storage module, and send the predicted abnormal data to the QoS policy decision module ;

QoS policy decision module: used to mark and store the abnormal data, report abnormal results, make a decision based on the abnormal data, and drive the execution of the decision.

Optionally, the data collection module specifically includes:

User terminal data acquisition unit: used to obtain the hardware data, software model version, installed applications, terminal location, moving direction, speed, CPU consumption, memory resources, and alarm logs of the user terminal;

Access network data acquisition unit: used to acquire base station distribution data, antenna channel mode, spectrum usage, physical resource and virtual resource usage, air interface signaling, and alarm log data;

Core network data collection unit: used to obtain user service quality agreements, network slice resource usage, core network signaling, and alarm log data;

Network KPI performance monitoring data collection unit: used to obtain network bandwidth, delay, and jitter data.

Optionally, the model training module uses a decision tree algorithm to construct the supervised machine learning model, and the result of the training is a tree-like structure consisting of nodes and branches, and each non-leaf node represents one of the data sets. Attribute, its branch is a certain value or value range of the attribute, and each leaf node is a category in the data set, indicating that the network service quality is abnormal or normal.

Optionally, the QoS policy decision module marks the abnormal results of the QoS abnormality monitor and the QoS abnormality predictor, stores new marked data in the QoS database, and updates the network in the QoS database. Quality of service data.

The invention proposed by the present invention proposes a 5G network service quality abnormal monitoring and prediction method, by collecting, storing, marking, and analyzing massive QoS service quality data of network terminals, wireless access networks, and core networks:

1) The relationship between historical network events and network service quality can be reconstructed;

2) Real-time monitoring of abnormal current network service quality, further forming network QoS management strategies, and automatic scheduling of network resources through software, virtualization, and slicing network programming interfaces, so as to ensure the quality of service for 5G network users, improve service quality;

3) It can predict possible abnormal network service quality in the future, and provide a basis for network planning and service quality optimization.

Description of drawings

In order to illustrate the technical solutions of the present invention more clearly, the following briefly introduces the accompanying drawings used in the technical description of the present invention. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention, which are of great significance to the art For those of ordinary skill, other drawings can also be obtained from these drawings without creative labor.

1 is a schematic flowchart of a method for monitoring and predicting abnormal network service quality provided by the present invention;

Fig. 2 is the marking format of user terminal data, access network data, core network data provided by the present invention;

3 is a schematic diagram of a decision tree for abnormal monitoring and prediction of network service quality provided by the present invention;

FIG. 4 is a schematic structural diagram of a network service quality abnormality monitoring and prediction system provided by the present invention.

Detailed ways

The present invention proposes a method and system for abnormal monitoring and prediction of 5G network service quality based on decision tree. The machine learning model realizes real-time monitoring and prediction of abnormal network service quality.

From the perspective of machine learning models, machine learning is divided into supervised learning, unsupervised learning and semi-supervised learning.

Under supervised learning, the input data is called "training data", and each set of training data has a clear identification or result, such as "spam" and "non-spam" in the anti-spam system, supervised learning Establish a learning process, compare the predicted results with the actual results of the "training data", and continuously adjust the prediction model until the predicted results of the model reach an expected accuracy rate. Supervised learning algorithms include linear regression, decision trees, Support Vector Machines, etc.

In unsupervised learning, the data is not specifically identified, and the model is learned to infer some intrinsic structure of the data. Common application scenarios include association rule learning and clustering, and unsupervised learning algorithms include K-means, hierarchical clustering, etc.

In the semi-supervised learning method, the input data is partially identified and partially unmarked. This learning model can be used to make predictions, but the model first needs to learn the internal structure of the data in order to organize the data reasonably to make predictions.

Machine learning mainly depends on algorithms, computing power and data. After data is obtained, it can be trained and calculated in an online or offline mode. The offline mode is not very real-time. Considering the real-time requirements of network service quality assurance, the present invention adopts an online method. way of learning and training.

In order to make the purpose, features and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the following The described embodiments are only some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

Referring to FIG. 1, the present invention proposes a method for monitoring and predicting abnormal 5G network service quality, the method includes:

S1. Collect 5G network service quality data and network KPI performance monitoring data, where the network service quality data includes user terminal data, access network data, and core network data;

The user terminal data includes: hardware data of the user terminal, software model version, installed applications, terminal location, moving direction, speed, consumed CPU, memory resources, and alarm logs;

The access network data includes: base station distribution, antenna channel mode, spectrum usage, usage of physical resources and virtual resources, air interface signaling, and alarm logs;

The core network data includes: user service quality agreement, network slice resource usage, core network signaling, and alarm logs;

The network KPI (Key Performance Indication) performance monitoring data includes: network bandwidth, delay, and jitter.

S2, preprocessing and marking the network service quality data;

The collected network service quality data is cleaned and formatted, and the network service quality data is marked in combination with the network KPI performance monitoring data, and the mark includes normal and abnormal.

Specifically, preprocessing is to remove noise and irrelevant data, correct errors, deal with invalid and missing values, uniformly normalize the collected data, and convert the data into a format that is easy for subsequent processing. Through network KPI performance monitoring data such as network bandwidth, delay, and jitter, the preprocessed user terminal data, access network data, and core network data are respectively marked as normal or abnormal. Some KPI performance data of the network can be monitored, such as rate, delay, etc., and the corresponding collected user data, access network data and core network data can be marked directly according to the monitored KPI data.

Please refer to FIG. 2. FIG. 2 lists the markup formats of user terminal data, access network data, and core network data. In Figure 2, Feature1, Feature2, etc. are attributes or features of the dataset, and the corresponding columns are the attribute values or value intervals of each item. There are two types of labeled category labels, Normal (normal) and Anomaly (abnormal). After marking, the marking rules can be learned and the prediction model can be trained through the marked network service quality data. Finally, the current network service quality can be automatically monitored and classified according to the learned rules to predict the results.

S3, storing the marked network service quality data in a QoS database;

Specifically, a relational database or NoSQL database or file system can be used for storage,

S4. Use the marked network service quality data in the QoS database as a data set, construct a supervised machine learning model, and use the data set to train the supervised machine learning model to obtain a QoS abnormality monitor and QoS anomaly predictor;

In the step S4, the supervised machine learning model adopts a decision tree algorithm. Algorithms that can be used also include neural network or support vector machine algorithms.

The decision tree algorithm is a supervised learning method for classification. The decision tree algorithm can infer decision rules from the characteristics of the training set data, and predict the result of the input data by creating a tree structure. Taking the marked network service quality data in the QoS database as a data set, the data set is generally divided into a training data set and a test data set, a decision tree is generated by the training data set, and how to predict the output is learned from the input data, Then use the test sample set to test, correct and repair the generated decision tree, verify the preliminary rules generated in the decision tree generation process through the data in the test data set, and prune those branches that affect the accuracy to generate more accurate Finally, the classification result (normal or abnormal) of the leaf node is used as the monitoring result of the newly input network service quality data.

Take the C4.5 decision tree algorithm as an example to illustrate the specific process of step S4:

1) Calculate the information gain rate of each attribute in the training data set, and build a decision tree model.

Specifically, the classification algorithm is based on information entropy. The larger the information entropy, the more new information the information carries, and the two data are significantly different, so they belong to different categories; the smaller the information entropy, the more new information that the information carries. The less, the two data are likely to belong to one class. If the value of the attribute is continuous, the continuous attribute is firstly discretized. After discretization, assuming that the attribute value of attribute A has m discrete value intervals, the training data set S is divided by the attribute value of attribute A into C ₁ , C ₂ , ..., C _m have m sub-data sets in total, |C _p | represents the number of samples in the p-th sub-data set, |S| represents the total number of samples in the data set before division,

Denote P(C _p ) as the frequency of the classification subset C _p in the sample set S, P(C _p )=|C _p |/|S|, p=1,2,...,m, the sample set before splitting The entropy of:

For any attribute A _i , suppose there are t different values a _q , q=1, 2,..., t, according to the different values of A _i , S can be divided into S ₁ , S ₂ ,..., S _t There are _{t subsets} in _total . At the same time, C ₁ , C ₂ , ..., C _m can be divided into m*t subsets. The entropy of the sample set after the attribute A _i is split:

in, The smaller the entropy, the higher the purity of the subset division. The information gain of the sample set after splitting by attribute A _i is:

InfoGain(S,A _i )=H(S)-H(S,A _i )

The information gain InfoGain(S,A _i ) represents the degree of uncertainty reduction after division.

Split information amount of attribute A _i :

Continue to split to create new nodes, and the information gain rate of the sample set S after splitting by attribute A _i is:

The C4.5 decision tree algorithm selects the attribute with the maximum information gain rate to build the initial decision tree from top to bottom, uses the test data set to prune the initial decision tree, removes abnormal branches, improves the classification accuracy, and obtains the final decision tree model.

Please refer to Fig. 3, a schematic diagram of anomaly monitoring and prediction decision tree, a decision tree is a tree structure like a flowchart composed of nodes and branches, wherein nodes are further divided into leaf nodes and non-leaf nodes. The top level of the tree is the first non-leaf node, the root node, which is the starting position of the decision tree. Each non-leaf node represents a certain attribute in the dataset, and its branch is a certain value or value range of the attribute. Each leaf node is a category in the dataset, which means that the network service quality is abnormal or normal.

2) The constructed decision tree module can be used as a QoS anomaly monitor, and the current 5G network service quality can be input into the decision tree model to monitor QoS anomalies; according to the current 5G network service quality data and historical 5G network services in the QoS data storage module The quality data can be further trained to obtain a QoS anomaly predictor to predict future 5G network service quality data anomalies.

S5, using the QoS abnormal monitor to monitor the current 5G network service quality data in real time, and send the monitored abnormal data to the QoS policy decision module;

S6. Use the QoS anomaly predictor to predict future 5G network service quality data anomalies, and send the predicted anomalous data to the QoS policy decision module;

In the step S6, the QoS anomaly predictor predicts future 5G network service quality data anomalies according to the current 5G network service quality data and historical 5G network service quality data, and the historical 5G network service quality data is in the QoS database. Saved records of historical network quality of service data.

S7. The QoS policy decision module marks and stores the abnormal data, updates the QoS database, reports the abnormal result, and makes a decision based on the abnormal data to drive the execution of the decision.

In the step S7, the QoS policy decision module marking and storing the abnormal data is specifically:

Automatically mark the results of the QoS anomaly monitor and the QoS anomaly predictor, store the new marked data in the QoS database, save it as historical 5G network service quality data, and update the network services in the QoS database quality data. In this way, the relationship between historical network events and network service quality is established, and the training basis is provided for the next training and prediction. After the QoS database is updated, when new network service quality data is input, the updated QoS database is used as a new data set to train and predict anomalies in the input data.

The QoS policy decision-making module also makes decisions based on the abnormal results monitored or predicted, such as increasing the bandwidth if the bandwidth is not enough, delaying too long, accelerating the queuing process, etc., and driving the decision-making execution.

Referring to FIG. 4, the present invention also provides a 5G network service quality abnormal monitoring and prediction system, the system includes:

Data collection module 410: used to collect 5G network service quality data and network KPI performance monitoring data, where the network service quality data includes user terminal QoS data, access network QoS data, and core network QoS data;

Data processing module 420: for preprocessing and marking the network service quality data;

QoS data storage module 430: used to store the marked network service quality data;

Model training module 440: used to use the marked network service quality data in the QoS database as a data set of a supervised machine learning model, construct a supervised machine learning model, and train the supervised machine learning model, Get QoS Anomaly Monitor and QoS Anomaly Predictor

QoS abnormality monitor 450: used to monitor the current 5G network service quality data in real time, and send the monitored abnormal data to the QoS policy decision module;

QoS abnormality predictor 460: used to predict the abnormality of future 5G network service quality data according to the current 5G network service quality data and the historical 5G network service quality data in the QoS data storage module, and send the predicted abnormal data to the QoS policy decision module;

QoS policy decision module 470: used to mark and store the abnormal data, report abnormal results, make a decision based on the abnormal data, and drive the execution of the decision.

The data collection module 410 specifically includes:

User terminal data acquisition unit: used to obtain the hardware data, software model version, installed applications, terminal location, moving direction, speed, CPU consumption, memory resources, and alarm logs of the user terminal;

Access network data acquisition unit: used to acquire base station distribution data, antenna channel mode, spectrum usage, physical resource and virtual resource usage, air interface signaling, and alarm log data;

Core network data collection unit: used to obtain user service quality agreements, network slice resource usage, core network signaling, and alarm log data;

Network KPI performance monitoring data collection unit: used to obtain network bandwidth, delay, and jitter data;

The model training module 440 adopts the decision tree algorithm to construct the supervised machine learning model, and the result of the training is a tree structure composed of nodes and branches, each non-leaf node represents an attribute in the data set, which A branch is a certain value or value range of the attribute, and each leaf node is a category in the data set, indicating that the quality of network service is abnormal or normal.

The QoS policy decision module 470 marks the abnormal results of the QoS abnormality monitor and the QoS abnormality predictor, stores the new marked data in the QoS database, and updates the network service quality data in the QoS database. .

The data collection module 410, the data processing module 420, the model training module 440, the QoS anomaly monitor 450, the QoS anomaly predictor 460, and the QoS policy decision module 470 together constitute a 5G network QoS machine learning engine. Access network QoS data, core network QoS data and network KPI performance monitoring data, using the 5G network QoS machine learning engine to automatically detect and predict network QoS anomalies, can further form network QoS management strategies to ensure service quality for 5G network users, It also provides a basis for network planning and network service quality optimization.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above may refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

In the foregoing embodiments, the description of each embodiment has its own emphasis. For parts that are not described or described in detail in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

Those of ordinary skill in the art can realize that the modules, units and/or method steps of various embodiments described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of the present invention.

As mentioned above, the above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand: The technical solutions described in the embodiments are modified, or some technical features thereof are equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (10)

1. A method for monitoring and predicting abnormal service quality of a 5G network is characterized by comprising the following steps:

s1, collecting 5G network service quality data and network KPI performance monitoring data, wherein the network service quality data comprises user terminal data, access network data and core network data;

s2, preprocessing the network service quality data and marking;

s3, storing the marked network service quality data to a QoS database;

s4, constructing a supervised machine learning model by taking the marked network service quality data in the QoS database as a data set, and training the supervised machine learning model by using the data set to obtain a QoS anomaly monitor and a QoS anomaly predictor;

s5, monitoring the current 5G network service quality data in real time by adopting the QoS anomaly monitor, and sending the monitored anomaly data to a QoS strategy decision module;

s6, predicting the future 5G network service quality data abnormity by adopting the QoS abnormity predictor, and sending the predicted abnormal data to a QoS strategy decision module;

and S7, the QoS strategy decision module marks and stores the abnormal data, updates a QoS database, reports an abnormal result, makes a decision according to the abnormal data and drives the decision to be executed.

2. The method for monitoring and predicting quality of service anomaly in 5G networks according to claim 1, wherein in step S1:

the user terminal data includes: hardware data and software model version of the user terminal, installed application, terminal position, moving direction, speed, consumed CPU, memory resource and alarm log;

the access network data includes: base station distribution, antenna channel mode, frequency spectrum usage, physical resource virtual resource usage, air interface signaling and alarm logs;

the core network data includes: user service quality agreement, network slice resource usage, core network signaling, and alarm log;

the network KPI performance monitoring data comprises: network bandwidth, delay, jitter.

3. The method for monitoring and predicting abnormal quality of service of 5G network according to claim 1, wherein the specific process of the step S2 is as follows:

and cleaning and unifying the format of the collected network service quality data, and marking the network service quality data by combining the network KPI performance monitoring data, wherein the marking comprises normal and abnormal.

4. The method for monitoring and predicting quality of service anomalies of a 5G network as claimed in claim 1 wherein, in said step S4, said supervised machine learning model employs a decision tree algorithm.

5. The method for monitoring and predicting quality of service anomaly of 5G network according to claim 1, wherein in step S6, the QoS anomaly predictor predicts future 5G network quality of service data anomaly according to current 5G network quality of service data and historical 5G network quality of service data, wherein the historical 5G network quality of service data is a historical network quality of service data record stored in the QoS database.

6. The method for monitoring and predicting quality of service (QoS) anomalies in a 5G network according to claim 1, wherein in the step S7, the QoS policy decision module marks and stores the anomaly data specifically as follows:

and automatically marking the results of the QoS anomaly monitor and the QoS anomaly predictor, storing new marking data into the QoS database, and updating network service quality data in the QoS database.

7. A 5G network quality of service anomaly monitoring and prediction system, the system comprising:

a data acquisition module: the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring 5G network service quality data and network KPI performance monitoring data, and the network service quality data comprises user terminal QoS data, access network QoS data and core network QoS data;

a data processing module: the system is used for preprocessing the network service quality data and marking the network service quality data;

a Qos data storage module: the system is used for storing the marked network service quality data;

a model training module: the system is used for constructing a supervised machine learning model by taking the marked network service quality data in the QoS database as a data set of the supervised machine learning model, and training the supervised machine learning model to obtain a QoS anomaly monitor and a QoS anomaly predictor

QoS anomaly monitor: the system is used for monitoring the current 5G network service quality data in real time and sending the monitored abnormal data to the QoS strategy decision module;

QoS anomaly predictor: the QoS policy decision module is used for predicting the future 5G network service quality data abnormity according to the current 5G network service quality data and the historical 5G network service quality data in the Qos data storage module and sending the predicted abnormal data to the QoS policy decision module;

a QoS policy decision module: the system is used for marking and storing the abnormal data, reporting an abnormal result, making a decision according to the abnormal data and driving the decision to execute.

8. The system for monitoring and predicting quality of service anomalies of a 5G network according to claim 7, wherein the data acquisition module specifically includes:

a user terminal data acquisition unit: the system comprises a Central Processing Unit (CPU), a memory resource and an alarm log, wherein the CPU, the memory resource and the alarm log are used for acquiring hardware data and software model versions of a user terminal, installed applications, terminal position, moving direction and speed, and consumed CPU, memory resource and alarm log;

an access network data acquisition unit: the method comprises the steps of acquiring base station distribution data, an antenna channel mode, frequency spectrum usage, physical resource virtual resource usage, air interface signaling and alarm log data;

a core network data acquisition unit: the system is used for acquiring user service quality agreement, network slice resource use, core network signaling and alarm log data;

the network KPI performance monitoring data acquisition unit: the method is used for acquiring network bandwidth, time delay and jitter data.

9. The system of claim 7, wherein the model training module employs a decision tree algorithm to construct the supervised machine learning model, the result of the training is a tree structure consisting of nodes and branches, each non-leaf node represents an attribute in the data set, a branch is a value or a range of values of the attribute, and each leaf node represents a category in the data set, and represents network quality of service anomaly or normality.

10. The system of claim 7, wherein the QoS policy decision module marks the QoS anomaly monitor and the QoS anomaly predictor, stores the new marked data in the QoS database, and updates the QoS data in the QoS database.