WO2022089031A1 - Network optimization method based on big data and artificial intelligence - Google Patents

Abstract

Disclosed is a network optimization method based on big data and artificial intelligence, relating to the technical field of network optimization and comprising: collecting multi-dimensional wireless network data and standard and customized signaling XDR data, performing data wrangling and cleaning by means of complementary fusion, and backfilling and storing longitudes and latitudes of user sampling points; acquiring precise user behavior by associating a displacement algorithm with a GIS architecture layer; determining a user-level moving state and stationary state by means of the displacement algorithm and indoor and outdoor user analysis, acquiring an indoor and outdoor user situation from signaling analysis, and modeling and calibrating quickly moving users and stationary indoor users based on the two types of information described above; and classifying cell users based on artificial intelligence and machine learning algorithms to form different sets of cells, randomly partitioning the sets of cells to train and validate a parameter learning model, and formulating a network optimization plan for the sets of cells by means of the validated parameter learning model. A network can be optimized and broadened by means of the present disclosure.

Description

A network optimization method based on big data and artificial intelligence

CROSS-REFERENCE TO RELATED APPLICATIONS

This application refers to Chinese Patent Application No. 2020111628946 filed on October 27, 2020, entitled "A Network Optimization Method Based on Big Data and Artificial Intelligence", which is fully incorporated into this application by reference.

technical field

The invention relates to the technical field of network optimization, in particular to a network optimization method based on big data and artificial intelligence.

Background technique

Traditional optimization and analysis methods based on network management data and field test data, as well as a single, general, and solidified optimization solution cannot meet the needs of complex customer perception improvement and network quality optimization in the 4G era. For some areas with large areas and sparsely populated areas, there are problems of complex network structure (difficulty in network planning and optimization), late start of 4G commercial use (there is a gap between 4G planning, optimization experience and advanced provinces), insufficient optimization personnel, and the original Some centralized optimization electronic processes lack effective automation and intelligent support means, which requires a large number of self-owned and manufacturer technical personnel to carry out planning and analysis work, and the optimization experience between different regions is different, and the optimization experience cannot be shared in time. Resulting in a high degree of dependence on manufacturers.

In response to the above problems, based on the research and application of big data and artificial intelligence, a network optimization method based on big data and artificial intelligence is designed and developed, which optimizes electronic online processes in a centralized manner, maximizes the network optimization rate, and supports maintenance staff to carry out optimization processing. Work.

SUMMARY OF THE INVENTION

The present invention provides a network optimization method based on big data and artificial intelligence to improve the efficiency of network planning, optimization and maintenance, and broaden and deepen the optimization work, aiming at the needs and deficiencies of current technology development.

A kind of network optimization method based on big data and artificial intelligence of the present invention, the technical scheme adopted to solve the above-mentioned technical problems is as follows:

A network optimization method based on big data and artificial intelligence, its implementation includes:

Collect MR-based multi-dimensional wireless network data and standard + customized signaling XDR data, and perform data sorting and cleaning through complementary integration to realize backfilling and storage of user sampling points in longitude and latitude;

Accurate user behavior is obtained through the association of displacement algorithm and GIS building layer;

Through the displacement algorithm and indoor and outdoor user analysis, the determination of the mobile state and the static state of the user level is realized. At the same time, the indoor and outdoor conditions of the user are obtained from the signaling analysis. Modelling and calibration of indoor users;

Based on the artificial intelligence machine learning algorithm, the cell users are classified to form different cell sets, and the parameter learning model is trained and verified by randomly dividing the cell sets. The verified parameter learning model can formulate different network optimization schemes based on different cell sets. .

Optionally, collect MR-based multi-dimensional wireless network data and standard + customized signaling XDR data. In this process, based on the APP used by the smartphone that can obtain location information, the data is collected through the Gb\IuPS\S1 interface. User data collection, by analyzing the latitude and longitude information reported by the user in the data signaling combined with the cell occupied by the signal, the user's fixed-point location can be obtained, and finally the data is collected and obtained.

Optionally, when a user uses an APP that can obtain location information through a smartphone, the user needs to call the data interface of assisted positioning, contain potential location information in the information exchange process, parse the relevant signaling and protocols, and transmit the user's location information. Order to extract and verify its accuracy, accurate available fields to achieve data storage.

Further optionally, after data sorting and cleaning is performed by means of complementary fusion, the MRO database table and the signaling XDR database table are obtained;

Realize the backfilling and storage of user sampling point latitude and longitude. In this process, it is necessary to model the historical wireless measurement information and known location information contained in MRO, and build the MRO database table and information based on the processing layer of Hadoop big data. Signal the correlation of the XDR database table, and then obtain the position of the user that only contains the wireless measurement information through the fingerprint database matching and reverse positioning, so as to realize the full position information of the full user and meet the requirements of position continuity.

Optionally, through the displacement algorithm and indoor and outdoor user analysis, the determination of the mobile state and the static state of the user level is realized. In this process:

First, label the user's motion state through the displacement algorithm to distinguish road user data and fixed-point user data;

Then, distinguish which users are high-speed mobile users by setting a time window, where:

(A) High-speed mobile user judgment: within the time window, the number of cell replacements is greater than or equal to a specified number of times, and the cell replacement distance is greater than the threshold distance;

(B) Low-speed mobile user judgment: within the time window, the number of cell replacements is less than the specified number or the distance between the first cell and the last cell is less than the threshold distance;

(C) User decision in static state: within the time window, the number of cell changes is equal to 0.

Further optionally, based on the artificial intelligence machine learning algorithm, the cell users are classified to form different cell sets, and the specific operations are:

Based on the ISODATA algorithm, set the threshold, classify according to the basic information of the cell configuration and the map, and obtain the geographical environment characteristics of the cell;

Based on the ISODATA algorithm, the threshold is set, and the collected data is classified to obtain the user service distribution characteristics;

Based on the collected data after sorting, optimization is performed to obtain optimized configuration characteristics related to the network;

Based on the collected data after cleaning, combined with network indicators and parameters, filter to obtain indicator characteristics;

The cell users are divided into different cell sets according to the four aspects of cell geographic environment characteristics, user service distribution characteristics, optimized configuration characteristics, and index characteristics.

Further optionally, the parameter learning model is trained and verified by randomly dividing the cell set. The verified parameter learning model can formulate different network optimization schemes based on different cell sets. The specific operations include:

Select any cell set, and randomly divide the data contained in the set into a training set and a test set by 7:3;

Use the GBDT algorithm to learn the training set and obtain the parameter learning model;

Use the test set to verify the validity of the parameter learning model. During the verification process, when the parameter learning model outputs the parameter configuration scheme that is most similar to the business needs of the community and has excellent index characteristics, the parameter learning model is considered to be effective. At this time, the parameter learning model outputs the The network optimization scheme of the cell, the network optimization scheme has the best parameter configuration.

Optionally, based on MR-based multi-dimensional wireless network data and standard + customized signaling XDR data, network problems can be precisely located, and point-line-plane network coverage of buildings, roads, and grids can be achieved, where:

(a) Take the building as a point to realize the overall coverage evaluation of the buildings in the whole network. On the one hand, the floors are stratified according to the building height information, and the MR data reported by the indoor resident users is used to realize the stratified coverage evaluation of the buildings. On the one hand, it can be accurate to the indoor floor of 10 meters, and comprehensively present the coverage and quality of the movement and competition in this area;

(b) With the road as the line, the macro-evaluation of the road index can be realized according to the positioning data at different times, and at the same time, the analysis of the moving and competing cells along the line can be carried out;

(c) Taking the outdoor part as the surface, according to the user MR data positioning results in the motion state, the outdoor area is covered and filled according to the 40*40 grid, so as to realize the outdoor coverage evaluation of the whole network and guide the site resource allocation;

Based on (a), (b), (c), according to the MR data collected after the frequency point measurement is turned on, combined with the MR indoor and outdoor positioning technology, the coverage of competition can be compared, and the mobile and telecom, mobile and China Unicom of a single building can be realized. The results of the coverage comparison are realized, and a comprehensive three-dimensional presentation is realized.

Compared with the prior art, a network optimization method based on big data and artificial intelligence of the present invention has the following beneficial effects:

1) The present invention transforms traditional research on a single network problem into a set of network problems that are related to each other, integrates network problems with a high degree of correlation and handles them, and formulates unique and targeted network optimization schemes for different types of problem cells, On the one hand, it can improve the efficiency of the original optimization work, and on the other hand, it can broaden and deepen the scope, means and content of the optimization work, and improve economic benefits;

2) The present invention can find abnormal problems of network structures such as longitude, latitude, direction angle, etc. in the network without manually going to the station, and provide a solution;

3) The multi-dimensional wireless network data and standard + customized signaling XDR data based on MR in the present invention can provide a means for network optimization to accurately locate network problems, and realize the point, line and surface network coverage of buildings, roads and grids The comprehensive presentation of , quality, competition and a series of optimization analysis applications support the network planning and optimization work in various provinces.

Description of drawings

FIG. 1 is a block diagram of a specific flow of execution (4) of Embodiment 1 of the present invention.

Detailed ways

In order to make the technical solutions of the present invention, the technical problems solved and the technical effects more clearly understood, the technical solutions of the present invention are described clearly and completely below with reference to specific embodiments.

Example 1:

This embodiment proposes a network optimization method based on big data and artificial intelligence, the implementation content of which includes:

(1) Collect MR-based multi-dimensional wireless network data and standard + customized signaling XDR data, and perform data sorting and cleaning through complementary integration to realize backfilling and storage of user sampling points in longitude and latitude.

(1.1) Collect MR-based multi-dimensional wireless network data and standard + customized signaling XDR data. In this process, based on the APP used by smartphones that can obtain location information, users can use the Gb\IuPS\S1 interface to perform For data collection, by analyzing the longitude and latitude information reported by the user in the data signaling combined with the cell occupied by the signal, the fixed-point location of the user can be obtained, and finally the data is collected and obtained.

When a user uses an APP that can obtain location information through a smartphone, he needs to call the data interface of assisted positioning, contain potential location information in the process of information exchange, parse the relevant signaling and protocols, extract the user's location information through signaling and verify it. Accuracy, accurate available fields to achieve data storage.

(1.2) Collect MR-based multi-dimensional wireless network data and standard + customized signaling XDR data, and after data sorting and cleaning through complementary fusion, the MRO database table and the signaling XDR database table are obtained.

(1.3) Realize the backfilling and storage of the longitude and latitude of user sampling points. In this process, it is necessary to model the historical wireless measurement information and known location information contained in the MRO, and build the MRO based on the processing layer of Hadoop big data. The correlation between the database table and the signaling XDR database table, and then reverse positioning through the fingerprint database matching to obtain the location of the user that only contains the wireless measurement information, to achieve the full amount of user location information, and to meet the requirements of location continuity.

(2) Accurate user behaviors can be obtained by associating displacement algorithms with GIS building layers.

(3) Through the displacement algorithm and indoor and outdoor user analysis, the determination of the mobile state and the static state of the user level is realized. At the same time, the indoor and outdoor conditions of the user are obtained from the signaling analysis, and the high-speed movement is realized through the combination of the above two key information. Modeling and calibration of users, stationary indoor users.

(3.1) Through the displacement algorithm and indoor and outdoor user analysis, the determination of the mobile state and the static state of the user level is realized. In this process:

(3.1.1) First, label the user's motion state through the displacement algorithm to distinguish road user data and fixed-point user data;

(3.1.2) Then, distinguish which users are high-speed mobile users by setting a time window, where:

(A) High-speed mobile user judgment: within the time window, the number of cell replacements is greater than or equal to a specified number of times, and the cell replacement distance is greater than the threshold distance;

(B) Low-speed mobile user judgment: within the time window, the number of cell replacements is less than the specified number or the distance between the first cell and the last cell is less than the threshold distance;

(C) User decision in static state: within the time window, the number of cell changes is equal to 0.

(4) Based on the artificial intelligence machine learning algorithm, the cell users are classified to form different cell sets, and the parameter learning model is trained and verified by randomly dividing the cell sets. The verified parameter learning model can be formulated based on different cell sets. Network optimization program.

The artificial intelligence machine learning algorithm can be the ISODATA algorithm. ISODATA is iterative self-organization analysis. By setting the initial parameters and using the mechanism of merging and splitting, when the distance between the centers of certain two types of clusters is less than a certain threshold, they are merged into one. Class, when the standard deviation of a class is greater than a certain threshold or the number of samples exceeds a certain threshold, it is divided into two categories. When the number of samples of a certain type is less than a certain threshold, it is canceled. In this way, according to the initial cluster center and the set number of categories and other parameters, an ideal classification result is finally obtained.

(4.1) Based on the artificial intelligence machine learning algorithm, the cell users are classified to form different cell sets. The specific operations are:

(4.1.1) Based on the ISODATA algorithm, set the threshold value, classify according to the basic information of the cell configuration and the map, and obtain the geographical environment characteristics of the cell;

(4.1.2) Based on the ISODATA algorithm, set the threshold, classify according to the collected data, and obtain the user service distribution characteristics;

(4.1.3) Based on the collected data, optimize it to obtain the optimized configuration characteristics related to the network;

(4.1.4) Based on the collected data after cleaning, combined with network indicators and parameters, filter to obtain indicator characteristics;

(4.1.5) Divide the cell users into different cell sets according to the four aspects of cell geographic environment characteristics, user service distribution characteristics, optimized configuration characteristics, and index characteristics.

(4.2) The parameter learning model is trained and verified by randomly dividing the cell set. The verified parameter learning model can formulate different network optimization schemes based on different cell sets. The specific operations include:

(4.2.1) Select any cell set, and randomly divide the data contained in the set into a training set and a test set by 7:3;

(4.2.2) Use the GBDT algorithm to learn the training set to obtain a parameter learning model;

(4.2.3) Use the test set to verify the validity of the parameter learning model. During the verification process, when the parameter learning model outputs the parameter configuration scheme that is most similar to the service requirements of the community and has excellent index characteristics, the parameter learning model is considered to be effective. At this time, The parameter learning model outputs the network optimization scheme of the cell, and the network optimization scheme has the best parameter configuration.

For this embodiment, it needs to be supplemented that: based on MR-based multi-dimensional wireless network data and standard + customized signaling XDR data, network problems can be accurately located, and point-line-plane network coverage of buildings, roads, and grids can be achieved. ,in:

(a) Take the building as a point to realize the overall coverage evaluation of the buildings in the whole network. On the one hand, the floors are stratified according to the building height information, and the MR data reported by the indoor resident users is used to realize the stratified coverage evaluation of the buildings. On the one hand, it can be accurate to the indoor floor of 10 meters, and comprehensively present the coverage and quality of the movement and competition in this area;

(b) With the road as the line, the macro-evaluation of the road index can be realized according to the positioning data at different times, and at the same time, the analysis of the moving and competing cells along the line can be carried out;

(c) Taking the outdoor part as the surface, according to the user MR data positioning results in the motion state, the outdoor area is covered and filled according to the 40*40 grid, so as to realize the outdoor coverage evaluation of the whole network and guide the site resource allocation;

Based on (a), (b), (c), according to the MR data collected after the frequency point measurement is turned on, combined with the MR indoor and outdoor positioning technology, the coverage of competition can be compared, and the mobile and telecom, mobile and China Unicom of a single building can be realized. The results of the coverage comparison are realized, and a comprehensive three-dimensional presentation is realized.

Taking Xinjiang mobile network problem analysis and optimization-Kuitun existing network as an example, a network optimization method based on big data and artificial intelligence proposed in this embodiment is described in detail. at this time:

According to (1) complete the collection of MR-based multi-dimensional wireless network data and standard + customized signaling XDR data, and organize and clean the data through complementary integration to realize backfilling and inputting of longitude and latitude of user sampling points library.

According to (2) the displacement algorithm is associated with the GIS building layer to obtain accurate user behavior.

According to (3) through the displacement algorithm and indoor and outdoor user analysis, the determination of the mobile state and the static state of the user level is realized. At the same time, the indoor and outdoor conditions of the user are obtained from the signaling analysis. Modeling and calibration of mobile users, stationary indoor users.

Referring to the attached drawings, according to (4) an artificial intelligence-based machine learning algorithm, taking into account the four aspects of the geographical environment characteristics of the cell, user service distribution characteristics, optimized configuration characteristics, and index characteristics, the cell users are classified to form different cell sets. A set of cells is used to train and verify the parameter learning model, and the verified parameter learning model can formulate different network optimization schemes based on different sets of cells.

What needs to be added is:

In the process of implementing (4), user service requirements can be analyzed based on experience. User service requirements are basically five categories: downlink video traffic, uplink video traffic, WeChat traffic, QQ communication traffic, and web browsing traffic. Therefore, the initialization clustering center can be Set to 2^5=32 categories (downlink video traffic (large, small), upstream video traffic (large, small)...), we consider the geographical environment characteristics of the cell, user service distribution characteristics, and optimized configuration characteristics based on the basic needs of business users. , The four aspects of index characteristics, complete the classification of users in the community.

In the process of implementing (4), different network optimization schemes are formulated for different cell sets based on the parameter learning model. Variance argumentation is carried out by comparing the rate and switching success rate to verify whether the network optimization scheme provided by the parameter learning model improves the network indicators.

The principles and implementations of the present invention are described in detail using specific examples above, and these examples are only used to help understand the core technical content of the present invention. Based on the above-mentioned specific embodiments of the present invention, those skilled in the art, without departing from the principles of the present invention, make any improvements and modifications made to the present invention, all should fall within the scope of patent protection of the present invention.

Claims (8)

1. A network optimization method based on big data and artificial intelligence, characterized in that its implementation content includes:

   Collect MR-based multi-dimensional wireless network data and standard + customized signaling XDR data, organize and clean data through complementary integration, and realize backfilling and storage of user sampling points in longitude and latitude;

   Accurate user behavior is obtained through the association of displacement algorithm and GIS building layer;

   Through the displacement algorithm and indoor and outdoor user analysis, the determination of the mobile state and the static state of the user level is realized. At the same time, the indoor and outdoor conditions of the user are obtained from the signaling analysis. Modelling and calibration of indoor users;

   Based on the artificial intelligence machine learning algorithm, the cell users are classified to form different cell sets, and the parameter learning model is trained and verified by randomly dividing the cell sets. The verified parameter learning model can formulate different network optimization schemes based on different cell sets. .
2. The method for network optimization based on big data and artificial intelligence according to claim 1, characterized in that, collecting multi-dimensional wireless network data based on MR and standard + customized signaling XDR data, in this process, Based on the APP used by smart phones that can obtain location information, user data is collected through the Gb\IuPS\S1 interface, and the latitude and longitude information reported by the user in the data signaling is analyzed and the cell is occupied to obtain the user's fixed-point location, and finally Acquisition of data.
3. A network optimization method based on big data and artificial intelligence according to claim 2, wherein when a user uses an APP that can obtain location information through a smartphone, he needs to call a data interface for assisted positioning, and during the information exchange process It contains potential location information, parses relevant signaling and protocols, extracts user location information from signaling and verifies its accuracy, and accurately uses fields to realize data storage.
4. A network optimization method based on big data and artificial intelligence according to claim 3, characterized in that, after data sorting and cleaning are carried out by means of complementary fusion, the MRO database table and the signaling XDR database table are obtained;

   Realize the backfilling and storage of user sampling point latitude and longitude. In this process, it is necessary to model the historical wireless measurement information and known location information contained in MRO, and build the MRO database table and information based on the processing layer of Hadoop big data. Signal the correlation of the XDR database table, and then obtain the position of the user that only contains the wireless measurement information through the fingerprint database matching and reverse positioning, so as to realize the full position information of the full user and meet the requirements of position continuity.
5. A kind of network optimization method based on big data and artificial intelligence according to claim 1, it is characterized in that, through displacement algorithm and indoor and outdoor user analysis, realize the determination of the mobile state and static state of the user level, in this process:

   First, label the user's motion state through the displacement algorithm to distinguish road user data and fixed-point user data;

   Then, by setting a time window to distinguish which users are high-speed mobile users,

   (A) High-speed mobile user judgment: within the time window, the number of cell replacements is greater than or equal to a specified number of times, and the cell replacement distance is greater than the threshold distance;

   (B) Low-speed mobile user judgment: within the time window, the number of cell replacements is less than the specified number or the distance between the first cell and the last cell is less than the threshold distance;

   (C) User decision in static state: within the time window, the number of cell changes is equal to 0.
6. A kind of network optimization method based on big data and artificial intelligence according to claim 1, it is characterized in that, based on artificial intelligence machine learning algorithm, classify cell users, form different cell sets, the specific operation is:

   Based on the ISODATA algorithm, set the threshold, classify according to the basic information of the cell configuration and the map, and obtain the geographical environment characteristics of the cell;

   Based on the ISODATA algorithm, the threshold is set, and the collected data is classified to obtain the user service distribution characteristics;

   Based on the collected data after sorting, optimization is performed to obtain optimized configuration characteristics related to the network;

   Based on the collected data after cleaning, combined with network indicators and parameters, filter to obtain indicator characteristics;

   The cell users are divided into different cell sets according to the four aspects of cell geographic environment characteristics, user service distribution characteristics, optimized configuration characteristics, and index characteristics.
7. A network optimization method based on big data and artificial intelligence according to claim 6, wherein the parameter learning model is trained and verified by randomly dividing the cell set, and the verified parameter learning model can be formulated based on different cell sets Different network optimization schemes, their specific operations include:

   Select any cell set, and randomly divide the data contained in the set into a training set and a test set by 7:3;

   Use the GBDT algorithm to learn the training set and obtain the parameter learning model;

   Use the test set to verify the validity of the parameter learning model. During the verification process, when the parameter learning model outputs the parameter configuration scheme that is most similar to the business needs of the community and has excellent index characteristics, the parameter learning model is considered to be effective. At this time, the parameter learning model outputs the The network optimization scheme of the cell, the network optimization scheme has the best parameter configuration.
8. A network optimization method based on big data and artificial intelligence according to claim 1, characterized in that, based on MR-based multi-dimensional wireless network data and standard + customized signaling XDR data, network problems can be precisely located , to realize the point-line-surface network coverage of buildings, roads, and grids, among which:

   (a) Take the building as a point to realize the overall coverage evaluation of the buildings in the whole network. On the one hand, the floors are stratified according to the building height information, and the MR data reported by the indoor resident users is used to realize the stratified coverage evaluation of the buildings. On the one hand, it can be accurate to the indoor floor of 10 meters, and comprehensively present the coverage and quality of movement and competition in this area;

   (b) Take the road as the line, realize the macro evaluation of the road index according to the positioning data at different times, and at the same time, analyze the moving and competing cells along the line;

   (c) Taking the outdoor part as the surface, according to the user MR data positioning results in the motion state, the outdoor area is covered and filled according to the 40*40 grid, so as to realize the outdoor coverage evaluation of the whole network and guide the site resource allocation;

   Based on (a), (b), (c), according to the MR data collected after the measurement of the frequency point competition is turned on, combined with the MR indoor and outdoor positioning technology, to realize the competitive comparison of coverage, and realize the mobile and telecom, mobile and China Unicom of a single building. Overlay the comparison results and achieve a comprehensive three-dimensional presentation.

Images