CN112689299A

CN112689299A - Cell load distribution method and device based on FP-growth algorithm

Info

Publication number: CN112689299A
Application number: CN202110049911.3A
Authority: CN
Inventors: 谢飞雄; 黄学彬; 余先保
Original assignee: BETTERCOMM
Current assignee: BETTERCOMM
Priority date: 2021-01-14
Filing date: 2021-01-14
Publication date: 2021-04-20
Anticipated expiration: 2041-01-14
Also published as: CN112689299B

Abstract

The invention relates to a cell load distribution method and a device based on an FP-growth algorithm, wherein the method comprises the following steps: carrying out prediction evaluation on the cell load; determining a working parameter association coverage grid data set according to the cell engineering parameters; determining an MRO coverage grid data set based on mass MRO measurement report data; processing the neighbor cell switching performance data; mining a worker parameter association coverage grid data set, an MRO coverage grid data set and processed neighbor cell switching performance data based on an FP-growth algorithm, and outputting a frequent data set and a strong association rule for load distribution; carrying out multi-dimensional comprehensive load distribution analysis processing to determine a final distribution cell list; and generating a load distribution scheme according to the distribution cell list and outputting the load distribution scheme. The method can fully automatically and quickly analyze various data, has high timeliness and quick and real-time scheme output; and through multi-dimensional comprehensive analysis and processing, the load shunting scheme is more accurate.

Description

Cell load distribution method and device based on FP-growth algorithm

Technical Field

The invention relates to the technical field of wireless communication, in particular to a cell load distribution method and a cell load distribution device based on an FP-growth algorithm.

Background

With the explosive growth of mobile terminals and network applications, the problem of shortage of wireless network resources is increasingly apparent. Meanwhile, due to the mobility of the user terminal and the imbalance of service partitions in different areas, a problem of high load of local network resources occurs, and the service experience of the user and the network performance of the terminal are seriously affected.

The load of network service is shunted, the service of a high-load cell coverage area is transferred to a wireless cell with relatively low load, the load of each wireless cell is balanced, and the performance of the whole network and the experience of users are improved while the local wireless cells are prevented from being overloaded. How to accurately identify a high-load cell in real time and accurately shunt traffic to a neighboring wireless cell is the core content of load shunting work.

The current solution mainly relies on manually collecting data such as network performance data, network configuration parameters, engineering parameters, etc., manually analyzing high-load and low-load wireless cells according to the experience of engineers, outputting a suitable shunting cell list, and manually adjusting related parameters or strategies to realize high-load cell shunting.

The prior method has the following defects:

1. the related data are manually collected/analyzed, the process is complicated, the timeliness is poor, and a real-time effective scheme cannot be output;

2. depending on the experience of engineers, the shunting scheme is easy to have misjudgment to cause the effect to be opposite;

3. the data source is single, which results in inaccurate splitting scheme.

Disclosure of Invention

In view of this, the present invention provides a cell load splitting method and device based on FP-growth algorithm to overcome the defects in the prior art.

In order to achieve the purpose, the invention adopts the following technical scheme: a cell load distribution method based on FP-growth algorithm includes:

carrying out prediction evaluation on the cell load so as to position a service high-load cell set;

determining a working parameter association coverage grid data set according to the cell engineering parameters;

determining an MRO coverage grid data set based on mass MRO measurement report data;

processing the neighbor cell switching performance data to convert the neighbor cell switching performance data into a data set supported by FP-growth;

mining a worker parameter association coverage grid data set, an MRO coverage grid data set and processed neighbor cell switching performance data based on an FP-growth algorithm, and outputting a frequent data set and a strong association rule for load distribution; wherein the frequent data set comprises: a working parameter coverage result set, an MRO coverage result set and a neighbor cell switching result set;

carrying out multi-dimensional comprehensive load distribution analysis processing on the load evaluation result set, the work parameter coverage result set, the MRO coverage result set and the neighbor cell switching result set to determine a final distribution cell list;

and binding a corresponding adjustment strategy to the final shunting cell list, and generating final adjustment instruction data as a load shunting scheme for output.

Optionally, the performing prediction evaluation on the cell load to locate the service high-load cell set includes:

collecting the performance statistical data of the wireless cell in the current fixed time in real time;

and based on LightGBM algorithm service load prediction, generating a service trend from massive historical load performance data by using a machine learning method, and determining the load state of the wireless cell by combining the current wireless cell performance statistical data.

Optionally, the determining the engineering parameter associated coverage grid data set according to the cell engineering parameter includes:

and performing geographic coverage simulation operation on the cell engineering parameters based on a GIS geographic information rasterization method to determine an engineering parameter associated coverage grid data set.

Optionally, the determining an MRO coverage grid data set based on the mass MRO measurement report data includes:

based on mass MRO measurement report data, extracting key data in the MRO measurement report data, converting the key data into map position information by using a specific algorithm, and determining a coverage grid data set.

Optionally, the determining a coverage grid data set based on the mass MRO measurement report data by extracting key data from the mass MRO measurement report data and converting the key data into map location information using a specific algorithm specifically includes:

carrying out invalid data cleaning on the MRO measurement report data to extract effective coverage data;

performing GIS geographic positioning processing on effective coverage data in MRO measurement report data;

and for the data after GIS geographic positioning processing, converting the longitude and latitude into corresponding grid identification, obtaining final rasterized data and forming a coverage grid data set.

Optionally, the processing the neighbor cell handover performance data to convert the neighbor cell handover performance data into a data set supported by FP-growth includes:

and converting the cell identification information in the neighbor cell switching performance data into a specific wireless cell object to form 2 data sets.

Optionally, the mining, working parameter association coverage grid data set, MRO coverage grid data set, and processed neighbor cell switching performance data based on the FP-growth algorithm, and outputting a frequent data set and a strong association rule for load distribution includes:

constructing a special strong association rule data mining module based on FP-growth;

and respectively inputting the work parameter association coverage grid data set, the MRO coverage grid data set and the processed adjacent cell switching performance data as input data to a special strong association rule data mining module, and outputting a corresponding frequent data set and a strong association rule through the processing of the special strong association rule data mining module.

Optionally, the processing procedure of the dedicated strong association rule data mining module includes:

establishing an item head table;

constructing an FP tree and generating a node linked list;

excavating a frequent item set of adjacent regions;

and (5) excavating a strong association rule of the adjacent regions.

Optionally, the performing multidimensional comprehensive load distribution analysis on the load evaluation result set, the working parameter coverage result set, the MRO coverage result set, and the neighboring cell switching result set to determine a final distribution cell list includes:

performing analysis processing on the shunting neighboring cells to extract an effective frequent neighboring cell relation list, and taking the frequent neighboring cell relation list as a primary shunting cell list;

2/3-layer neighbor cell cleaning is carried out on the preliminary shunting cell list to obtain a two-step shunting cell list;

and carrying out load distribution priority weighting on the two-step distribution cell list to obtain a final distribution cell list and a distribution priority.

The invention also provides a cell load distribution device based on FP-growth algorithm, comprising:

the load evaluation module is used for carrying out prediction evaluation on the cell load so as to position a service high-load cell set;

the first determining module is used for determining a project parameter association coverage grid data set according to the cell project parameters;

the second determining module is used for determining an MRO coverage grid data set based on the mass of MRO measurement report data;

the conversion module is used for processing the neighbor cell switching performance data so as to convert the neighbor cell switching performance data into a data set supported by FP-growth;

the mining module is used for mining the worker parameter association coverage grid data set, the MRO coverage grid data set and the processed neighbor cell switching performance data based on the FP-growth algorithm, and outputting a frequent data set and a strong association rule for load distribution; wherein the frequent data set comprises: a working parameter coverage result set, an MRO coverage result set and a neighbor cell switching result set;

the multi-dimensional comprehensive processing module is used for carrying out multi-dimensional comprehensive load distribution analysis processing on the load evaluation result set, the work parameter coverage result set, the MRO coverage result set and the neighbor cell switching result set to determine a final distribution cell list;

and the load distribution scheme output module is used for binding a corresponding adjustment strategy to the final distribution cell list and generating final adjustment instruction data as the load distribution scheme to be output.

By adopting the technical scheme, the cell load distribution method based on the FP-growth algorithm carries out load prediction based on LightGBM algorithm service, generates a service trend from massive historical load performance data by using a machine learning method, and determines the load state of a wireless cell by combining with the current load performance data; performing geographic coverage simulation operation on the cell engineering parameters based on a rasterization method of GIS geographic information processing to determine an associated coverage grid data set; extracting key data based on mass MRO measurement report data, converting the key data into map position information by using a specific algorithm, and determining a coverage grid data set; mining neighbor cell switching performance data, cell parameter association coverage grid data and MRO coverage grid data based on an FP-growth algorithm, and outputting a frequent data set and a strong association rule scheme for load distribution; based on the multidimensional data analysis of 4 types of result data (a load evaluation result set, an engineering parameter coverage result set, an MRO coverage result set and a neighbor cell switching result set), a precise load distribution list is generated by adopting a special data analysis algorithm, and a load distribution scheme is generated and output according to the load distribution list. The method can fully automatically and quickly analyze various data, has high timeliness and faster and real-time scheme output; and based on performance load data, engineering parameter geography coverage simulation and MRO measurement data coverage analysis, multi-dimensional comprehensive analysis processing is carried out on the neighbor cell switching performance data, so that the load distribution scheme is more accurate.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

FIG. 1 is a schematic diagram of an overall flow provided by a FP-growth algorithm-based cell load distribution method according to the present invention;

FIG. 2 is a schematic diagram of a geographical dotting of a cell location;

FIG. 3 is a schematic diagram of geographic area rasterization;

fig. 4 is a schematic diagram of wireless cell coverage main lobe and tail lobe area calculation;

FIG. 5 is a schematic diagram of GIS processing MRO data;

FIG. 6 is an overall view of FP-growth mining of neighbor handover performance data;

FIG. 7 is a schematic diagram of data mining;

FIG. 8 is a schematic diagram of the construction of an FP-tree;

FIG. 9 is a node chain table;

FIG. 10 is the condition base of CELL 3;

FIG. 11 is a multi-dimensional integrated split processing schematic;

FIG. 12 is a schematic diagram of a frequent set determination offload principle;

FIG. 13 is a schematic diagram of a strong association rule for weighting a load shedding list;

FIG. 14 is a schematic illustration of a split strategy generation adjustment instruction scheme output;

fig. 15 is a schematic view of an overall flow provided by the FP-growth algorithm-based cell load splitting method according to the present invention.

In the figure: 1. a load evaluation module; 2. a first determination module; 3. a second determination module; 4. a conversion module; 5. a digging module; 6. a multi-dimensional comprehensive processing module; 7. and a load distribution scheme output module.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be described in detail below. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. All other embodiments, which can be derived by a person skilled in the art from the examples given herein without any inventive step, are within the scope of the present invention.

Fig. 1 is a schematic view of an overall flow provided by a cell load splitting method based on an FP-growth algorithm according to the present invention.

As shown in fig. 1, a cell load splitting method based on FP-growth algorithm according to the present invention includes:

s11: carrying out prediction evaluation on the cell load so as to position a service high-load cell set;

s12: determining a working parameter association coverage grid data set according to the cell engineering parameters;

s13: determining an MRO coverage grid data set based on mass MRO measurement report data;

s14: processing the neighbor cell switching performance data to convert the neighbor cell switching performance data into a data set supported by FP-growth;

s15: mining a worker parameter association coverage grid data set, an MRO coverage grid data set and processed neighbor cell switching performance data based on an FP-growth algorithm, and outputting a frequent data set and a strong association rule for load distribution; wherein the frequent data set comprises: a working parameter coverage result set, an MRO coverage result set and a neighbor cell switching result set;

s16: carrying out multi-dimensional comprehensive load distribution analysis processing on the load evaluation result set, the work parameter coverage result set, the MRO coverage result set and the neighbor cell switching result set to determine a final distribution cell list;

s17: and binding a corresponding adjustment strategy to the final shunting cell list, and generating final adjustment instruction data as a load shunting scheme for output.

The cell load distribution method carries out deep mining on massive neighbor cell switching performance data, MRO measurement report data and cell parameter data through an FP-growth frequent item set mining algorithm, carries out multi-dimensional distribution strategy analysis by combining cell load state prediction data, and outputs an optimal distribution scheme of a high-load wireless cell.

The specific implementation of the method is described in detail below.

1. And predicting and evaluating the cell load, and mainly used for positioning a service high-load cell set. By collecting the wireless cell performance statistical data within a fixed time (such as 15 minutes) in real time, generating a service trend from massive historical load performance data by using a machine learning method based on LightGBM algorithm service load prediction, determining the load state of the wireless cell by combining the current wireless cell performance statistical data, accurately positioning a high/low load area, and carrying out quantitative grading on cell loads.

The load prediction adopts a light GBM (light Gradient Boosting machine) algorithm, and is mainly based on a decision tree algorithm. The LightGBM is similar to the XGBoost and is also a gradient lifting framework, but different from the XGBoost, it selects leaf growth (each layer only branches one node), and using the histogram algorithm only needs to find the splitting point after ordering all data at the beginning, and only needs to simply perform the bucket splitting operation each time the splitting point is found. All the groups are exhausted according to a greedy algorithm, and the basis for finding the optimal segmentation point is to obtain the optimal solution and a target value according to a first derivative and a second derivative. The LightGBM algorithm has the advantages of faster training memory usage, higher accuracy, support for parallelized learning, and capability of processing large-scale data.

The training data set can be used for predicting and selecting various data, and the load evaluation prediction mainly relates to the following KPI (Key Performance indicator), which is shown in a table 1:

TABLE 1

And determining the load level of the current wireless cell by comprehensively predicting and judging the plurality of KPI performance indexes. Table 2 shows a logic determination method of the load level:

table 2 table 3 exemplifies the load class output data:

TABLE 3

2. Geographical area information rasterized data creation

Geographic area rasterization is the projection onto a geographic plane using the geographic information of all the cell data and rasterizing the entire map in a 10 m by 10 m area square grid.

As shown in fig. 2, after the longitude and latitude of the cell are dotted through a map, the longitude and latitude of a point a (upper left point) and the longitude and latitude of a point B (lower right point) of the known map data are determined, and then a grid coordinate system is determined and established every 10 meters from the longitude and latitude directions by taking the point a as a starting point until the coordinate system just contains the longitude and latitude of the point B, and grid operation is ended.

With the established grid coordinate system, a grid data set is established from the X-axis (longitude) of the a point (upper left point) as a starting point. As shown in FIG. 3, all grid points are listed in sequence number by X1Y1-XnYn, outputting all grid point data sets.

3. GIS geographic association coverage rasterization data set based on engineering parameters

By inputting engineering parameters (longitude and latitude, azimuth solution, cell type and the like) of the wireless cell, GIS simulation operation is carried out on geographic space coverage, a correlation coverage grid data set of the wireless cell is determined, coverage correlation on a geographic position can be determined through further deep excavation, and a shunting cell list based on the engineering parameter correlation coverage is determined.

3.1 cell coverage area grid set

Through longitude and latitude and direction angles in wireless cell engineering, a grid set related to a sector area of a cell main lobe and a cell tail lobe horizontal coverage direction is solved by using a two-point distance algorithm and a two-point azimuth angle algorithm.

As shown in fig. 4, the coverage distances in the main lobe/tail lobe directions of the cell are D1 and D2, the horizontal coverage angles of the main lobe/tail lobe of the cell are Ia1 and Ia2, the azimuth angle is Az, and the latitudes and longitudes of the cell are lon1 and lat1, respectively.

And the planning calculation methods of the adjacent cells are different for wireless cells with different coverage types. The most common wireless base stations are macro base stations and micro base stations, and the carrier power of the macro base stations is much higher than that of the micro base stations, and the coverage capacity is also much higher. Macro base stations are generally used for outdoor coverage, micro base stations are used for indoor coverage, and the antenna types are distinguished according to the most common antenna types of wireless cells: the macro base station adopts a directional antenna, and the micro base station adopts an omnidirectional antenna.

Key parameters for different coverage scenarios and cell types are as shown in table 4:

TABLE 4

3.2 method for solving grid set of macro cell main lobe

Step 1: coverage start/end azimuth solution:

step 2: and acquiring a grid set RL10 with the geographic distance to the wireless cell being less than or equal to D1 in all grid sets by using distance calculation.

And step 3: the azimuth angle operation is used to calculate a grid set RL11 whose azimuth angle from the own radio cell to the center point of the grid is [ Az10, Az11] in the grid set RL 10.

3.3 method for solving grid set of macrocell tail lobes

Step 1: solving the tail lobe azimuth Az':

Az'＝Mod(180+Az,360)

step 2: coverage start/end azimuth solution:

and step 3: and acquiring a grid set RL20 with the geographic distance to the wireless cell being less than or equal to D2 in all grid sets by using distance calculation.

And 4, step 4: the azimuth angle operation is used to calculate a grid set RL21 whose azimuth angle from the own radio cell to the center point of the grid is [ Az0, Az1] in the grid set RL 20.

3.4 grid set solving method of microcell tail lobe

Specifically, a distance calculation is used for obtaining a grid set RL31 with the geographic distance from the wireless cell being less than or equal to D1 in all grid sets.

3.5 GIS geographic operation method

In the above solution, the GIS geographic operation methods include distance operation based on two-point longitude and latitude and direction angle calculation based on two-point longitude and latitude. The specific calculation method is as follows:

1) distance calculation of wireless cell from midpoint of grid

The calculation method is supposed to work out as a perfect sphere, and the earth surface distance between any two points on the earth surface can be calculated according to the longitude and latitude of the two points. If the longitude and latitude of the wireless cell are Lon1 and Lat1, respectively, and the longitude and latitude of the middle point of the grid are Lon2 and Lat2, respectively, the calculation formula of the distance D between two points based on the spherical cosine formula is:

D＝R*arccos(cosLat1)*cos(Lat2)*cos(Lon2-Lon1)+sin(Lat1)*sin(Lat2)

wherein: r is the earth radius of 6378.137 km; the distance D is in units of kilometers.

2) Wireless cell to grid midpoint azimuth calculation

Since the direction angle calculation has directivity, the azimuth calculation is an azimuth from the wireless cell to the center point of the grid unidirectionally. The longitude and latitude of the wireless cell are respectively Lon1 and Lat1, the longitude and latitude of the middle point of the grid are respectively Lon2 and Lat2, and the solution formula of the azimuth solution A simplified by longitude and latitude into plane geometry is as follows:

wherein: the point B is in the first quadrant and the positive half shaft of the Y axis, and Azimuth is equal to A;

b is in the second quadrant, Azimuth is 360+ A;

b is in the third four quadrants and Y axis negative half axis, Azimuth is 180+ A.

3.6 Geographically associating the output of the coverage grid set

The data format of the cell coverage area grid set calculated by the above method is data using a cell as a primary key, as shown in table 5:

table 5 example data for updating to a grid set with statistically relevant overlay cells in grid number, as shown in table 6:

TABLE 6

4. Coverage grid dataset based on MRO measurements

And performing conversion by using key information in the MRO data, dotting the data to a map position, updating the associated coverage cells in the data points to a geographic grid set after data-free cleaning of the data points, and performing deep data mining to obtain a flow distribution neighbor list based on the MRO.

The MR measurement report is one of the most frequent data from the UE (user equipment) in the mobile radio network system, and is the measurement of the UE on the radio network environment quality in the current state, and the MR measurement data is periodically stored in the form of an MRO file after being processed by the mobile radio coverage base station eNodeB. The main data involved are: timing advance TA, antenna arrival angle AOA and RSRP/RSRQ value of the measuring cell.

As shown in table 7, MRO data main fields and meanings:

TABLE 7

4.1, invalid data scrubbing

Data validity mainly extracts data with good wireless quality, and eliminates useless quality difference data to determine effective coverage data. And extracting the data with good wireless quality, and judging whether the wireless quality measurement of the serving cell and the adjacent cell meets the requirement by using ScRSRP, ScRSRQ, NcRSRP and NcRSRQ. The following conditions need to be satisfied for good wireless quality:

RSRP > -110dB and RSRQ > -15dB

And substituting the ScRSRP/ScRSRQ of the main serving cell and the NcRSRP/NcRSRQ of the adjacent cells into the formula to extract a data set with good wireless quality.

4.2, carrying out GIS geographic positioning processing on MRO data

The latitude and longitude of the cell can be determined by using the information of the main service cell in the MRO data, the distance between the time advance TA and the service cell can be calculated, and the position of the UE reporting the MRO data can be determined by combining the antenna arrival angle AOA.

And searching an external cell information table of the corresponding base station through the eNBId, and determining a longitude value and a latitude value in specific engineering parameters of the service cell by combining the ScEarfcn and the ScPci. And then the ScAOA antenna reaching angle is used for determining the longitude and latitude values of the corresponding MRO data packet. The principle of GIS processing MRO data is shown in figure 5.

In fact, the GIS operator module mainly performs the transformation of the reaching angle of the ScAOA antenna, the calculation of the scadv distance, and the solution of the longitude and latitude of the second point, and outputs the MRO coverage result data shown in table 8 by using the serving cell and the neighboring cell after the GIS geographic positioning processing is performed on the MRO data:

TABLE 8

4.3 overlay grid data output

In the MRO coverage result data output by the above process, we can convert the longitude and latitude into a corresponding grid identifier, and obtain final rasterized data, an example of which is shown in table 9:

TABLE 9

5. Neighbor handover performance dataset processing

The neighbor cell switching performance data reflects the service switching situation between the wireless cell statistics and the neighbor cells, and can initially position the inflow and outflow situations of services. The neighbor cell switching performance data needs to be converted into an FP-growth available data set, and a frequent data set and a strong association rule are mined from the FP-growth available data set. The FP-growth mining process of the neighbor handover performance data is shown in fig. 6.

The mining of the neighbor cell switching performance data mainly comprises the following 3 steps:

step 1: data conversion

The neighbor cell handover performance data only includes the cell identification information (sECI, nceci), and the information needs to be converted into a specific wireless cell object, so as to form 2 data sets.

Step 2: item header table establishment

Because the adjacent cell switching performance data is the switching times of every two cells, the time counting data needs to generate an item head table according to rules.

And step 3: FP-growth data mining

Inputting the information of the '2 item data sets' and the 'item head table' obtained in the last two steps into a special strong association rule data mining module based on FP-growth, and obtaining a related data mining result: 2/3/4 item frequent set, strong association neighborhood rule.

6. Frequent set and strong association mining analysis of FP-growth

6.1, constructing a special strong association rule data mining module based on FP-growth

And respectively carrying out data mining on the geographic association coverage grid data set, the MRO coverage grid data set and the processed neighbor cell switching performance data set by utilizing an FP-growth algorithm principle, and outputting strong association rules of the three types of data, namely high-confidence association rule result data.

Fig. 7 is a schematic diagram of the data mining module, taking the employee reference coverage grid data set as an example, and the specific principle flow is as follows.

6.1.1 input data example

In section 4.3, the processed results are shown in Table 10:

mro overlay data set

Watch 10

6.1.2 establishing item head table

The item head table records the occurrence frequency of all 1 item frequent sets, eliminates the data sets with low support degree (< ═ 20%), and arranges the data sets in descending order according to the frequency, as shown in table 11:

head table

TABLE 11

TABLE 12

6.1.3 building FP Tree and generating node chain table

And generating an item head table and the sorted data set to start the construction of the FP tree. And starting to read in the sorted data set one by taking null as a root node, inserting the data set into the FP tree, and inserting the data set into the FP tree according to the sorted sequence. And inserting the nodes into the FP tree, wherein the nodes in the front of the sequence are ancestor nodes, and the nodes in the back are descendant nodes. If there is a common ancestor, the corresponding common ancestor node count is incremented by 1. After insertion, if a new node appears, the node corresponding to the entry head table is linked with the new node through the node linked list. And completing building the FP tree until all data are inserted into the FP tree. FIG. 8 shows the construction of FP-tree.

When the FP tree is constructed, related nodes can be connected by combining the item head table to generate a node linked list. The resulting linked list is shown in FIG. 9.

6.1.4 mining frequent item sets of adjacent regions

After the FP tree, the item head table and the node linked list are obtained, the bottom items of the item head table are mined upwards in sequence. For each entry of the entry header table corresponding to the FP-tree, its conditional pattern base is found. The conditional mode base is to take the node to be mined as the FP sub-tree corresponding to the leaf node, set the count of each node in the sub-tree as the count of the leaf node, delete the node with the count lower than the support degree, and then recursively mine to obtain the frequent item set of the adjacent region.

Taking the bottommost Cell3 as an example, as shown in fig. 10, the condition pattern bases are { Cell1:2, Cell2:1, Cell4:1}, { Cell1:2, and Cell3:1}, and after removing the low-

support cells

2 and 4, the final frequent 2 sets are: { Cell1:2, Cell3:2 }.

In the invention, the maximum use of 4 item sets is required as a judgment basis. After the operation, the result of the frequent set of all nodes (i.e. neighbors) is shown in table 13:

watch 13

6.1.5 outputting neighbor strong association rule

And (3) mining the strong association rules of the adjacent regions, firstly constructing all possible 2 rules by using an adjacent region relation table, then calculating the confidence coefficient of each rule one by one, and outputting all the rules with the confidence coefficients larger than the minimum confidence coefficient. Example data of 2 strong association rules for a neighbor cell are shown in table 14:

serving cell	Neighboring cell	Confidence level
			Cell1	Cell2	0.8
Cell1	Cell3	0.6
			Cell1	Cell4	0.3
Cell1	Cell5	0.1

TABLE 14

7. Multidimensional comprehensive load split analysis

After data processing in a number of steps, a total of 4 types of resulting data sets are obtained: a load evaluation result set, an engineering parameter coverage result set, an MRO coverage result set and an adjacent cell switching result set. And performing multi-dimensional comprehensive analysis processing on the 4 types of data to obtain a final shunting cell list. Fig. 11 is a schematic diagram of multidimensional comprehensive analysis processing.

7.1 result data set

The result data set is processed input data, and is respectively derived from:

load evaluation result set: section 1 output result of prediction and evaluation of cell load, mainly load grade data of wireless cell;

covering a result set by the worker parameters: 3.6, the data of the geographical association coverage grid set output by the subsection passes through 2, 3 and 4 frequent sets and 2 strong association neighbor cell rules deeply mined by FP-growth;

MRO coverage result set: 4.3 MRO coverage grid data set data output by the subsection pass through 2, 3 and 4 frequent sets and 2 strong association adjacent region rules deeply mined by FP-growth;

neighbor cell switching result set: and 2, 3 and 4 frequent sets and 2 strongly-associated neighbor cell rules are deeply mined from neighbor cell performance switching data output by the 6 sections through FP-growth.

7.2 Split neighborhood analysis processing

The method comprises the steps of using a high-load cell in a load evaluation result set to obtain a neighbor cell configuration table to construct 2 data sets, and extracting an effective frequent neighbor cell relation list by combining 2 frequent sets of three types of data, namely an engineering parameter coverage result set, an MRO coverage result set and a neighbor cell switching result set, wherein the list is called a primary shunting cell list.

As shown in fig. 12, 2 frequent sets { Cell2, Cell3}, { Cell3, and Cell4} are obtained through extraction, which indicates that Cell2 and Cell3 or Cell3 and Cell4 are effective shunting neighbor relations, that is, Cell2 and Cell4 may be selected as shunting target cells when Cell3 is in a high load state.

7.3, 2/3 level neighborhood cleaning

2/3 level neighbor cell cleaning mainly aims at the preliminary shunting cell list to carry out further shunting effectiveness judgment. Because the current load state can be determined without direct performance data of the 2-layer adjacent cell (adjacent cell) or the 3-layer adjacent cell, whether the possibility of service inflow and outflow exists between high-load wireless cells can be determined through the 3/4 frequent sets (corresponding to the 2-layer/3-layer adjacent cells). And by combining the load state in the load evaluation result set, the high-load shunting cells of the 2-layer/3-layer adjacent cells can be removed, so that the problem of high load of the target cell caused by improper shunting is avoided.

As shown in fig. 12, as it is known that { Cell2, Cell3, Cell4} is 3 frequent sets, it can be determined that the traffic flow direction correlation between Cell2 and Cell4 is very strong, and at this time, if Cell4 is a high-load Cell, the splitting scheme of { Cell2- > Cell3} needs to be removed, so that it can avoid load increase of Cell4 due to the fact that the split traffic from Cell2 flows to Cell4 through Cell 3.

Through cleaning the 2/3-layer neighbor cell, a two-step shunting cell list can be more accurately output.

7.4 Strong Association rule weighting

The weighting of the strong association rule mainly comprises the steps of carrying out load distribution priority weighting on a two-step distribution cell list, and weighting the distribution cells by using confidence coefficients in 2 strong association rules in three types of data of an engineering parameter coverage result set, an MRO coverage result set and a neighboring cell switching result set, so as to form a final distribution cell list and distribution priority. The principle of weighting the load shedding list by the strong association rule is shown in fig. 13.

8. Load split plan output

After the shunting scheme of the multidimensional comprehensive analysis is processed, a load shunting scheme list is obtained, namely, a corresponding adjustment strategy can be bound to a load shunting target cell, and final adjustment instruction data is generated and output as a scheme. FIG. 14 is a schematic illustration of a split strategy generation adjustment instruction scheme output.

In fig. 14, "own cell" and "diverting cell" in the "diverting parameter setting object" correspond to "serving cell" and "neighboring cell" in the "final diverting cell list", respectively;

"shunt parameter type": this field may correspond to a specific adjustment instruction;

whether the branching priority setting is supported or not indicates that the parameter type can set different parameter values for the branching cells according to different branching priorities.

The cell load distribution method based on the FP-growth algorithm carries out load prediction based on LightGBM algorithm service, generates service trend from massive historical load performance data by using a machine learning method, and determines the load state of a wireless cell by combining with the current load performance data; performing geographic coverage simulation operation on the cell engineering parameters based on a rasterization method of GIS geographic information processing to determine an associated coverage grid data set; extracting key data based on mass MRO measurement report data, converting the key data into map position information by using a specific algorithm, and determining a coverage grid data set; mining neighbor cell switching performance data, cell parameter association coverage grid data and MRO coverage grid data based on an FP-growth algorithm, and outputting a frequent data set and a strong association rule scheme for load distribution; based on the multidimensional data analysis of 4 types of result data (a load evaluation result set, an engineering parameter coverage result set, an MRO coverage result set and a neighbor cell switching result set), a precise load distribution list is generated by adopting a special data analysis algorithm, and a load distribution scheme is generated and output according to the load distribution list. The method can fully automatically and quickly analyze various data, has high timeliness and faster and real-time scheme output; and based on performance load data, engineering parameter geography coverage simulation and MRO measurement data coverage analysis, multi-dimensional comprehensive analysis processing is carried out on the neighbor cell switching performance data, so that the load distribution scheme is more accurate.

As shown in fig. 15, a cell load splitting device based on FP-growth algorithm according to the present invention is characterized in that:

the load evaluation module 1 is used for predicting and evaluating the cell load so as to position a service high-load cell set;

the first determining module 2 is used for determining a project parameter association coverage grid data set according to the cell project parameters;

the second determining module 3 is configured to determine an MRO coverage grid data set based on a large amount of MRO measurement report data;

the conversion module 4 is configured to process the neighbor cell handover performance data to convert the neighbor cell handover performance data into a data set supported by FP-growth;

the mining module 5 is used for mining worker parameter association coverage grid data sets, MRO coverage grid data sets and processed neighbor cell switching performance data based on an FP-growth algorithm, and outputting frequent data sets and strong association rules for load distribution; wherein the frequent data set comprises: a working parameter coverage result set, an MRO coverage result set and a neighbor cell switching result set;

the multidimensional comprehensive processing module 6 is used for carrying out multidimensional comprehensive load distribution analysis processing on the load evaluation result set, the work parameter coverage result set, the MRO coverage result set and the neighbor cell switching result set to determine a final distribution cell list;

and the load distribution scheme output module 7 is configured to bind a corresponding adjustment strategy to the final distribution cell list, and generate final adjustment instruction data to be output as a load distribution scheme.

The working principle of the cell load splitting device based on the FP-growth algorithm is the same as that of the cell load splitting method based on the FP-growth algorithm, and the working principle is not repeated herein.

It is understood that the same or similar parts in the above embodiments may be mutually referred to, and the same or similar parts in other embodiments may be referred to for the content which is not described in detail in some embodiments.

It should be noted that the terms "first," "second," and the like in the description of the present invention are used for descriptive purposes only and are not to be construed as indicating or implying relative importance. Further, in the description of the present invention, the meaning of "a plurality" means at least two unless otherwise specified.

Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing specific logical functions or steps of the process, and alternate implementations are included within the scope of the preferred embodiment of the present invention in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the present invention.

It should be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.

It will be understood by those skilled in the art that all or part of the steps carried by the method for implementing the above embodiments may be implemented by hardware related to instructions of a program, which may be stored in a computer readable storage medium, and when the program is executed, the program includes one or a combination of the steps of the method embodiments.

In addition, functional units in the embodiments of the present invention may be integrated into one processing module, or each unit may exist alone physically, or two or more units are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may also be stored in a computer readable storage medium.

The storage medium mentioned above may be a read-only memory, a magnetic or optical disk, etc.

In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.

Claims

1. A cell load distribution method based on FP-growth algorithm is characterized by comprising the following steps:

2. The method of claim 1, wherein the performing a predictive assessment of cell load to locate a traffic high load cell set comprises:

3. The cell load splitting method according to claim 1, wherein the determining the working parameter associated coverage grid data set according to the cell engineering parameters comprises:

4. The cell load splitting method according to claim 1, wherein the determining an MRO coverage grid data set based on a mass of MRO measurement report data includes:

5. The cell load splitting method according to claim 4, wherein the determining a coverage grid data set by extracting key data based on the mass of MRO measurement report data and converting the key data into map location information using a specific algorithm specifically comprises:

6. The cell load splitting method according to claim 1, wherein the processing the handover performance data of the neighboring cell to convert the handover performance data into the FP-growth supported data set includes:

7. The cell load splitting method according to claim 1, wherein the mining work parameter association coverage grid data set, MRO coverage grid data set, and processed neighbor cell handover performance data based on FP-growth algorithm, and outputting frequent data set and strong association rule for load splitting includes:

8. The cell load splitting method according to claim 7, wherein the processing procedure of the dedicated strong association rule data mining module includes:

establishing an item head table;

constructing an FP tree and generating a node linked list;

excavating a frequent item set of adjacent regions;

and (5) excavating a strong association rule of the adjacent regions.

9. The cell load splitting method according to claim 1, wherein the performing multidimensional comprehensive load splitting analysis on the load evaluation result set, the working parameter coverage result set, the MRO coverage result set, and the neighbor cell switching result set to determine a final split cell list comprises:

10. A cell load splitting device based on FP-growth algorithm is characterized by comprising: