WO2016188498A1

WO2016188498A1 - Wireless network throughput evaluating method and device

Info

Publication number: WO2016188498A1
Application number: PCT/CN2016/084549
Authority: WO
Inventors: 顾军; 张兴; 易正磊
Original assignee: 中兴通讯股份有限公司
Priority date: 2015-10-21
Filing date: 2016-06-02
Publication date: 2016-12-01
Also published as: CN106612511A; CN106612511B

Abstract

A wireless network throughput evaluating method and device, comprising: acquiring historical throughput data of N base stations; according to the obtained historical throughput data of the N base stations, constructing a base station association network of the N base stations; according to the constructed base station association network and the obtained historical throughput data, finding M base stations important for the evaluation of a throughput of the base station association network, and considering the M base stations as important base stations; using the historical throughput data of the determined M base stations, evaluating a throughput of the remaining (N-M) base stations. In embodiments of the present invention, historical data of base station throughputs of selected important base stations is used to evaluate other base stations, thus reducing data complexity.

Description

Method and device for evaluating wireless network throughput

Technical field

This document relates to, but is not limited to, data mining technology, and in particular to a method and device for evaluating wireless network throughput.

Background technique

With the rapid development of wireless networks, the types and traffic of mobile Internet data services have been greatly improved. The explosive growth of traffic and the extremely rich types of services have made the analysis of network traffic behavior more complicated.

In order to effectively implement network planning and design, network resource allocation, and refined operation management, it is necessary to accurately analyze network throughput. Due to the diversity, randomness and suddenness of data services, the data analysis methods in related technologies are too complicated to meet the current network throughput behavior analysis.

Summary of the invention

The following is an overview of the topics detailed in this document. This Summary is not intended to limit the scope of the claims.

Embodiments of the present invention provide a method and an apparatus for evaluating wireless network throughput, which can satisfy current network throughput behavior analysis.

According to an aspect of an embodiment of the present invention, a method for evaluating a wireless network throughput is provided, including:

Obtaining historical data of throughput of N base stations;

Constructing a base station relationship network of the N base stations according to the acquired historical data of the throughput of the N base stations;

Obtaining M base stations that play an important role in evaluating the throughput of the base station relationship network according to the constructed base station relationship network and the historical data of the obtained throughput, and using the M base stations as important base stations;

Using the historical data of the determined throughput of the M important base stations, the remaining N-M Base station throughput is evaluated;

Where N and M are both positive integers and N is greater than M.

Optionally, the historical data of acquiring the throughput of the N base stations includes:

Collecting an original throughput sequence of each base station, and calculating an average value of the throughput sequence values in the collected original throughput sequence;

The value of the sequence of the preset percentage thresholds in which the sequence value is prior to each other in the original throughput sequence of each base station is replaced by the calculated average of the throughput sequence values. Value, get the new throughput sequence for each base station;

The normalized throughput sequence of each base station is obtained by normalizing the time series of the new throughput of each base station, and the obtained normalized throughput sequence is used as the historical data of the throughput.

Optionally, the constructing the base station relationship network of the N base stations according to the historical data of the acquired throughput of the N base stations includes:

Calculating correlation coefficients between two of the N base stations according to the normalized throughput sequence of each of the obtained base stations;

When the calculated correlation coefficient is greater than a correlation coefficient threshold, an undirected edge is generated between the two base stations;

Constructing the base station relationship network of the N base stations by using an undirected edge generated between two base stations of the N base stations.

Optionally, the M base stations that use the constructed base station relationship network and the acquired historical data of the throughput to find an important effect on the base station relationship network throughput evaluation performance include:

Obtaining the degree of each base station by counting the number of undirected edges of each base station in the base station relationship network;

Selecting the base stations of the N most basic base stations in sequence, and evaluating the throughput of the N-m base stations according to the support vector machine, and obtaining the throughput evaluation effect of the m base station relationship networks;

In the obtained throughput evaluation effect of the m base station relationship networks, the best throughput evaluation effect is selected, and the m base stations corresponding to the selected best throughput evaluation effects are used as the base stations. M base stations that play an important role in the network throughput evaluation effect;

Where m, M, and N are positive integers, M<=m, M<N, and m<N.

Optionally, the number of undirected edges of each base station is proportional to the size of the base station.

Optionally, the estimating the remaining N-M base station throughputs by using the determined historical data of the throughput of the M important base stations includes:

Constructing a throughput relationship model between the remaining N-M base stations and M important base stations by using a support vector machine algorithm;

The estimated throughput of the remaining N-M base stations is obtained by using the constructed throughput relationship model and the throughput history data of the M important base stations.

According to another aspect of the embodiments of the present invention, an apparatus for evaluating wireless network throughput is provided, including:

An acquisition module, configured to collect historical data of throughput of N base stations;

a building module, configured to construct a base station relationship network of the N base stations according to the acquired historical data of the throughput of the N base stations;

a searching module, configured to use the constructed base station relationship network and the acquired historical data of the throughput to find M base stations that play an important role in evaluating the throughput of the base station relationship network, and use the M base stations as important base stations ;

An evaluation module, configured to use the historical data of the determined throughput of the M important base stations to evaluate the remaining N-M base station throughputs;

Where N and M are both positive integers and N is greater than M.

Optionally, the collecting module includes:

Calculating a throughput average unit, configured to collect a raw throughput sequence of each base station, and calculate an average value of the throughput sequence values in the collected original throughput sequence;

The obtaining unit is configured to replace the calculated throughput with a sequence of a preset percentage threshold of the sequence value before each sequence of the original throughput sequence of each of the acquired base stations is replaced by the calculated The average of the sequence values, the new throughput sequence for each base station, and the normalization of the time series of the new throughput for each base station, resulting in each The normalized throughput sequence of the base station, the resulting normalized throughput sequence is taken as historical data of the throughput.

Optionally, the building module includes:

Calculating the correlation coefficient unit is configured to calculate correlation coefficients between two of the N base stations according to the normalized throughput sequence of each of the obtained base stations;

Generating the undirected edge unit to set an undirected edge between the two base stations when the calculated correlation coefficient is greater than the correlation coefficient threshold;

The building unit is configured to construct the base station relationship network of the N base stations by using an undirected edge generated between two base stations of the N base stations.

Optionally, the searching module includes:

The obtaining unit is configured to obtain the degree of each base station by counting the number of undirected edges of each base station in the base station relationship network, and sequentially select the m base stations with the medium most middle of the N base stations, according to the support vector machine Assessing the throughput of the Nm base stations, and obtaining the throughput evaluation effect of the m base station relationship networks;

The searching unit is configured to select the best throughput evaluation effect in the throughput evaluation effect of the obtained m base station relationship networks, and use the m base stations corresponding to the selected best throughput evaluation effect as the base station M base stations that play an important role in the network throughput evaluation effect;

Where m, M, and N are positive integers, M<=m, M<N, and m<N.

Compared with the related art, the technical solution provided by the embodiment of the present invention includes: acquiring historical data of throughput of N base stations; and constructing a base station relationship network of N base stations according to the acquired historical data of throughput of the N base stations According to the constructed base station relationship network and the historical data of the acquired throughput, find M base stations that play an important role in the base station relationship network throughput evaluation effect, and use the M base stations as important base stations; use the determined M important Historical data of the throughput of the base station, and the remaining NM base station throughput is evaluated. The beneficial effects of the embodiments of the present invention are as follows: the embodiment of the present invention uses the historical data of the base station throughput of a selected number of important base stations to evaluate the characteristics of other base stations, thereby reducing the complexity of data analysis; When the throughput of the base part of the base station is known, the throughput of other unknown base stations in the spatial range is evaluated, thereby providing a reference for optimizing the wireless network resources.

Other aspects will be apparent upon reading and understanding the drawings and detailed description.

BRIEF abstract

FIG. 1 is a flowchart of a method for evaluating wireless network throughput according to an embodiment of the present invention;

2 is a schematic diagram of an apparatus for evaluating wireless network throughput according to an embodiment of the present invention;

3 is a flowchart of a method for evaluating a wireless network throughput according to an embodiment of the present invention;

4 is a flowchart of an algorithm of a support vector machine according to an embodiment of the present invention;

FIG. 5 is a network diagram for constructing a base station relationship according to a first embodiment of the present invention; FIG.

6 is a diagram showing the variation of the average value of SMAPE (Symmetric mean absolute percentage error) with m according to the first embodiment of the present invention;

7 is a diagram showing evaluation results of two base stations according to a first embodiment of the present invention;

8 is a network diagram for constructing a base station relationship according to a second embodiment of the present invention;

Figure 9 is a diagram showing the variation of the average value of SMAPE with m according to the second embodiment of the present invention;

FIG. 10 is a diagram showing evaluation results of two base stations according to a second embodiment of the present invention.

Embodiments of the invention

Embodiments of the present application will be described in detail below with reference to the accompanying drawings. It should be noted that, in the case of no conflict, the features in the embodiments and the embodiments in the present application may be arbitrarily combined with each other.

FIG. 1 is a flowchart of a method for evaluating throughput of a wireless network according to an embodiment of the present invention. As shown in FIG. 1 , the method includes the following steps:

Step 101: Acquire historical data of throughput of N base stations;

It should be noted that, in the embodiment of the present invention, N may include the number of base stations in the area where the wireless network throughput is evaluated.

Step 102: Construct a base station relationship network of N base stations according to the acquired historical data of the throughput of the N base stations.

Step 103: Using the constructed base station relationship network and the historical data of the acquired throughput, find M base stations that play an important role in evaluating the throughput of the base station relationship network, and use the M base stations as important base stations;

Step 104: Evaluate the remaining N-M base station throughputs by using historical data of the determined throughput of the M important base stations;

Where N and M are both positive integers and N is greater than M.

Optionally, obtaining historical data of the throughput of the N base stations includes: collecting an original throughput sequence of each base station, and calculating an average value of the throughput sequence values in the collected original throughput sequence; The value of the sequence of the preset percentage thresholds in which the sequence value is prior to each sequence in the original throughput sequence of each base station is replaced by the average of the calculated throughput sequence values, and the new throughput of each base station is obtained. The sequence of normalization throughput is obtained by normalizing the sequence of the new throughput of each base station, and the normalized throughput sequence obtained is used as the historical data of the throughput.

It should be noted that, in the embodiment of the present invention, the preset percentage threshold may be 3%, and 3% is only an optional value. The preset percentage threshold is a value obtained by a person skilled in the art according to empirical analysis, and may be 2% to 5%. A value in . The original throughput data may be collected by using the original throughput sequence of the base station for all the durations, or the original throughput sequence of the preset duration. The preset duration is generally greater than or equal to 14 days when the original throughput sequence of the preset duration is used.

The base station relationship network for constructing the N base stations according to the historical data of the acquired throughput of the N base stations includes: calculating, according to the normalized throughput sequence of each base station obtained, two base stations of the N base stations respectively. Correlation coefficient; when the correlation coefficient is greater than the correlation coefficient threshold, an undirected edge is generated between the two base stations; the base station relationship of the N base stations is constructed by the undirected edges generated between the two base stations in the N base stations The internet.

It should be noted that calculating the correlation coefficient between the two base stations is a common technical means by those skilled in the art, and the correlation coefficient threshold may be 0.6.

Optionally, using the constructed base station relationship network and the historical data of the acquired throughput, the M base stations that play an important role in evaluating the throughput of the base station relationship network include: performing statistics on the undirected side of each base station in the base station relationship network. The number of the nodes is obtained, and the degree of each base station is obtained; the m base stations with the most moderate base stations are sequentially selected, and the throughput of the Nm base stations is evaluated according to the support vector machine, and the throughput evaluation effect of the m base station relationship networks is obtained; Among the throughput evaluation effects of the m base station relationship networks, the best throughput evaluation effect is selected, and the m base stations corresponding to the selected best throughput evaluation effects are regarded as important for the base station relationship network throughput evaluation effect. M a base station; wherein m, M, and N are positive integers, M<=m, M<N, and m<N. In the embodiment of the present invention, the number of undirected edges is proportional to the size of the base station, and such a proportional relationship is common knowledge of those skilled in the art.

The utilization of the throughput history data of the M important base stations to evaluate the remaining NM base station throughput includes: constructing a throughput relationship model between the remaining NM base stations and the M important base stations by using a support vector machine algorithm; The relationship model and the throughput history data of the M important base stations obtain the estimated throughput of the remaining NM base stations.

It should be noted that the construction of the throughput relationship model is a common technical means by those skilled in the art. In addition to the construction of the support vector machine algorithm in the embodiment of the present invention, other algorithms may be used to determine the throughput relationship model. After constructing the throughput relationship model, the throughput history data of the M base stations is taken as an input, and the evaluation throughput of the N-M base stations can be output.

The embodiment of the invention further provides a computer storage medium, wherein the computer storage medium stores computer executable instructions, and the computer executable instructions are used to perform the above-mentioned wireless network throughput evaluation method.

2 is a schematic diagram of an apparatus for evaluating wireless network throughput according to an embodiment of the present invention. As shown in FIG. 2, the method includes: an acquisition module 201, a construction module 202, a lookup module 203, and an evaluation module 204. The collecting module 201 is configured to acquire historical data of the throughput of the N base stations; the constructing module 202 is configured to construct a base station relationship network of the N base stations according to the acquired historical data of the throughput of the N base stations; the searching module 203 is configured In order to utilize the constructed base station relationship network and the historical data of the acquired throughput, find M base stations that play an important role in evaluating the throughput of the base station relationship network, and use the M base stations as important base stations; the evaluation module 204 is configured to utilize The historical data of the throughput of the M important base stations is determined, and the remaining NM base station throughputs are evaluated; wherein N and M are both positive integers, and N is greater than M.

Optionally, the collecting module 201 includes: a calculating throughput average unit, configured to collect an original throughput sequence of each base station, and calculate an average value of the throughput sequence values in the collected original throughput sequence;

The obtaining unit is configured to replace the calculated throughput sequence with the value of the sequence of the preset percentage threshold of the sequence value before each sequence of the original throughput sequence of each base station collected The average of the values, the new throughput sequence of each base station is obtained, and the normalized throughput sequence of each base station is obtained by normalizing the time series of the new throughput of each base station, and the obtained normalized throughput sequence is obtained. The throughput sequence is used as historical data for throughput.

The constructing module 202 includes: calculating a correlation coefficient unit configured to calculate a correlation coefficient between two base stations of the N base stations according to the normalized throughput sequence of each obtained base station;

The generating undirected edge unit is set to generate an undirected edge between the two base stations when the calculated correlation coefficient is greater than the correlation coefficient threshold;

The building unit is configured to construct a base station relationship network of N base stations by using undirected edges generated between two base stations of the N base stations.

The lookup module 203 includes:

The obtaining unit is configured to obtain the degree of each base station by counting the number of undirected edges of each base station in the base station relationship network, and sequentially select m base stations with the most moderate among the N base stations, and evaluate Nm base stations according to the support vector machine. Throughput, the throughput evaluation effect of m base station relationship networks is obtained;

The searching unit is configured to select the best throughput evaluation effect in the throughput evaluation effect of the obtained m base station relationship networks, and use the m base stations corresponding to the best throughput evaluation effect as the base station relationship. M base stations that play an important role in network throughput evaluation;

Where m, M, and N are positive integers, M<=m, M<N, and m<N.

The embodiment of the invention mainly comprises the following four modules: a data preprocessing module (equivalent to an acquisition module), a base station relationship network construction module (equivalent to a construction module), an important base station selection module (equivalent to a search module), and a space throughput evaluation module. (Equivalent to the evaluation module). The data pre-processing module is configured to select the N base stations to be studied and eliminate the abnormal data points; the base station relationship network construction module is configured to construct the base station relationship between the base stations according to the historical data of the throughput of the collected N base stations. The network; the important base station selection module is configured to select M important base stations from the N base stations according to the constructed historical data of the base station relationship network and the throughput; the space throughput evaluation module is set to be used for the time to be evaluated, according to the known The determined throughput of the M base stations evaluates the throughput of the other NM base stations.

The data preprocessing module can be configured to include:

A1. Select N base stations in the same area in the spatial location;

A2. Excluding the abnormal point in the historical data of the throughput of each base station; here, the abnormal point includes the point where the sequence value is extremely large, that is, the value of the sequence of the preset percentage threshold after the sequence value is sorted, and the sequence value is before, Excluding the anomaly point includes replacing the value of the sequence of the outliers with the average of the calculated throughput sequence values.

A3. Normalize the data once.

The base station relationship network building module can be configured to include:

B1. Calculating a correlation coefficient between two base stations of the N base stations;

B2. Based on the correlation coefficient, construct a base station relationship network with a given correlation coefficient threshold.

The important base station selection module can be set to include:

C1. Counting the size of each base station in the base station relationship network;

C2. sequentially select the base station with M (M=1, 2...N) before the degree as the important base station, and evaluate the throughput of the other N-M base stations according to the support vector machine;

C3. Select the M base stations that evaluate the best effect (acting role) on the historical data of the throughput as important base stations.

The space throughput assessment module includes:

D1. Evaluate the throughput of other N-M base stations according to the selected M important base stations.

3 is a flowchart of a method for evaluating a wireless network throughput according to an embodiment of the present invention. As shown in FIG. 3, the method includes:

Step 1, data preprocessing;

In order to evaluate the throughput of other large base stations according to the throughput of some base stations in the future, historical data of the throughput of all base stations needs to be acquired first, and then the acquired historical data is preprocessed. Data preprocessing mainly consists of the following steps:

a, selecting N base stations in the same area at the spatial location according to requirements;

b. arranging the original throughput sequences of the N base stations, and replacing the values of the sequence of the preset percentage thresholds in the sequence of the sequence values before each sequence of the throughput sequence with the average value of the calculated throughput sequence values. For example, replacing the first 3% of the throughput in the throughput sequence with the average of the sequence values of the throughput sequence to obtain the throughput sequence p ⁱ (i=1, 2...N) of the i-th base station, As a new throughput sequence;

c. normalize each base station throughput sequence to obtain a normalized throughput sequence S _{i of} the i-th base station, and use the obtained normalized throughput sequence as historical data of throughput.

among them,

For the normalized throughput sequence at time t of the i-th base station, max(p _i ),min(p _i ) represent the maximum and minimum values of the original throughput sequence, respectively, and L is the total length of the sequence of the throughput sequence.

Step 2: construct a base station relationship network;

For the N base stations to be studied, L is the total duration of the collected data (where the data is the throughput sequence), taking L before the T (general

Calculating the correlation coefficient ρ _ij between the i-th (i=1, 2, 3...N) base stations and the jth (j=1, 2, 3...N) base stations by using the time data of the left and right)

S _i is the throughput sequence of the i-th base station,

The average throughput of the i-th base station over the total duration,

The throughput of the i-th base station at time t (t=1, 2, 3...T); S _j is the throughput sequence of the j-th base station,

The average throughput of the jth base station over the total duration,

The throughput of the jth base station at time t (t = 1, 2, 3 ... T). For a given correlation coefficient threshold c, if ρ _{ij is} greater than c, it is considered that there is a significant correlation between base station i and base station j, and an undirected edge is added between them, so that the base station relationship of N base stations can be constructed. The internet.

Step 3. Select an important base station;

In the present invention, a support vector machine (SVM) is used to evaluate base station throughput. The algorithm flow of the SVM is shown in Figure 4 and includes the following steps:

1. Establish a training sample set and a test sample set according to the evaluation sample;

2. Establish an objective function according to the training sample set;

3. Solve the objective function and get the optimal parameters;

4. Substituting the optimal parameters into the objective function to obtain a decision regression equation;

5. Verify the decision regression equation using test data;

6, whether it is less than the given error e;

When it is judged that it is smaller than the given error e, the process proceeds to step 7, and when it is judged that it is not smaller than the given error e, the parameter is adjusted, and the process returns to step 3.

7. Calculate the sample input decision regression equation to calculate the throughput of other base stations.

In the present invention, the final throughput evaluation effect is measured by the symmetric mean relative error (SMAPE), and the SMAPE reflects the relative error between the evaluation value and the real value, and solves the problem that the real value is too small. The problem of relative error is too large, the formula is:

Where Ft is the evaluation value and At is the actual value.

In order to select some base stations as important base stations, the top m (m=1, 2...N) base stations with the largest degree are selected as the important base stations, that is, M base stations that play an important role in evaluating the throughput of the base station relationship network, according to the support. The vector machine evaluates the throughput of other Nm base stations. Calculate the SMAPE of each base station that is evaluated, and select the M base stations with the lowest average SMAPE as the important base stations.

Step 4. Use other important base stations to evaluate other base station throughput.

In the embodiment of the present invention, according to the selected M important base stations, the SVM algorithm is used, and the throughput relationship model of other N-M base stations and M important base stations is trained using the historical data of the throughput. The throughput of the M base stations in the time period to be evaluated is input into the relational model, and the throughput of the corresponding N-M base stations can be output.

Embodiments of the present invention will be described below with reference to FIGS. 5 to 10.

Embodiment 1

The data in this example is derived from the statistics of all base stations in a large city. The time granularity is 60 minutes and the total length of time is 21 consecutive days. The wireless network space throughput evaluation method in the embodiment of the present invention includes the following steps:

Step 1: Data preprocessing;

A. Select 95 base stations in the same area at the spatial location according to the requirements;

B. Excluding the abnormal data points in 95 base stations, and obtaining the throughput sequence of each base station;

C. Normalize each base station throughput sequence to obtain a normalized throughput sequence S _{i for} the i-th base station.

Step 2: Construct a base station relationship network for 95 base stations to be studied;

A. Calculate the correlation between the i-th (i=1, 2, 3...95) base stations and the jth (j=1, 2, 3...95) base stations by taking the data of the first 18 days of the 95 base stations. The coefficient ρ _{ij is} calculated as

Where T=432, S _i is the throughput sequence of the i-th base station,

The average throughput of the i-th base station over the total duration,

The throughput of the i-th base station at time t (t=1, 2, 3...432); S _j is the throughput sequence of the j-th base station,

The average throughput of the jth base station over the total duration,

The throughput of the jth base station at time t (t = 1, 2, 3 ... 432).

B. In the embodiment of the present invention, a given correlation coefficient threshold c=0.6 (it is generally considered that the correlation coefficient is greater than 0.6 is a strong correlation), and if ρ _{ij is} greater than 0.6, an undirected edge is added between the base station i and the base station j. So, you can build a network of 95 base stations. As shown in FIG. 5, the dot represents the base station, and the undirected edge reflects the correlation between the base stations. The larger the point, the greater the degree of the base station; wherein, the larger the point, the more the undirected side, the undirected side The more, the greater the degree.

Step 4: Select an important base station;

A. From the 18-day historical data, the first 15 days of data is selected as the training sample set, and the last 3 days of data is used as the test sample set; the first 15 days of data of all base stations is used as the input of the support vector machine algorithm (SVM), and the output training is obtained. The other 95-m base stations and the selected m base station throughput relationship models;

B. The last 3 days of the m base stations with the largest degree are used as the input of the throughput relationship model, and the estimated values of the other 3 days after the other 95-m base stations are output;

C. Calculate the corresponding SMAPE of each base station of 95-m base stations, and make the average value of SMAPE The change of m, as shown in Figure 6, the black point is the average value of 95-m base station SMAPE. It can be seen from the figure that when m=8, the average SMAPE of other base stations is the smallest, which is the best prediction effect. Therefore, the number of important base stations selected in this embodiment is M=8.

Step 5: Use the support vector machine algorithm to estimate the space throughput.

In the embodiment of the present invention, according to the selected eight important base stations, the support vector machine algorithm (SVM) uses the historical data of the throughput to train the throughput relationship model of the other 87 base stations and the 8 important base stations. By inputting the throughput of the eight base stations in the last three days of the original 21-day data into the relational model, the throughput of the corresponding 87 base stations can be output.

As shown in FIG. 7, the evaluation results of some of the 87 base stations are shown, where 1 is the evaluation value and 2 is the real value. Calculating the evaluation error of 87 base stations, and obtaining an average SMAPE=30.3%, it can be seen that the method of the embodiment of the present invention has high accuracy.

Embodiment 2

The data in this example is derived from the statistical data of a typical area in a large city with a time granularity of 60 minutes and a total length of time of 18 consecutive days. The wireless network space throughput evaluation method in the embodiment of the present invention includes the following steps:

Step 1: Data preprocessing;

A. Select 117 base stations in the same area at the spatial location according to the requirements;

B. Excluding the abnormal data points in 117 base stations, and obtaining the throughput sequence of each base station;

Step 2: 117 base stations to be studied, and construct a base station relationship network;

A. Calculate the correlation between the i-th (i=1, 2, 3...117) base stations and the jth (j=1, 2, 3...117) base stations by taking the data of the first 15 days of the 117 base stations. The coefficient ρ _{ij is} calculated as

Where T=360, S _i is the throughput sequence of the i-th base station,

For the average throughput of the i-th base station over the total duration,

The throughput of the i-th base station at time t (t=1, 2, 3...360); S _j is the throughput sequence of the j-th base station,

The average throughput of the jth base station over the total duration,

The throughput of the jth base station at time t (t = 1, 2, 3 ... 360).

B. In the embodiment of the present invention, a given correlation coefficient threshold c=0.6 (it is generally considered that the correlation coefficient is greater than 0.6 is a strong correlation), and if ρ _{ij is} greater than 0.6, an undirected edge is added between the base station i and the base station j. So, you can build a network of 117 base stations. As shown in FIG. 8 , the point represents the base station, and the correlation between the base stations is reflected. The greater the point, the greater the degree of the base station; wherein, the larger the point, the more the undirected side, the more the undirected side The greater the degree.

Step 4: Select an important base station;

A. From the 15 days of historical data, the first 12 days of data are selected as the training set, and the last three days of data are used as the test set; the first 12 days of data from all base stations are used as input to the support vector machine algorithm (SVM), and the output training is obtained. a throughput relationship model between 117-m base stations and selected m base stations;

B. The last 3 days of the m base stations with the largest degree are used as the input of the throughput relationship model, and the estimated values of the other 117-m base stations are outputted for the next three days;

C. Calculate the average SMAPE change with m for each SMAPE of each base station of 117-m. As shown in Figure 9, the black point is the average value of 117-m base station SMAPE, as can be seen from the figure. When m=11, the average SMAPE of other base stations is the smallest, that is, the prediction effect is the best, so in this embodiment, the number of important base stations we select is M=11.

Step 5: Evaluate space throughput using the SVM algorithm.

In the embodiment of the present invention, according to the selected 11 important base stations, the support vector machine algorithm (SVM) uses the historical data of the throughput to train the throughput relationship model of the other 106 base stations and 11 important base stations. By inputting the throughput of 11 base stations in the time period to be evaluated into the relational model, the throughput of the corresponding 106 base stations can be output.

As shown in FIG. 10, it is an example of the evaluation result, where 3 is an evaluation value and 4 is a true value. The evaluation error of 106 base stations is calculated to obtain an average SMAPE=36.4%, and the evaluation result of the embodiment of the present invention has high accuracy.

In summary, the embodiments of the present invention have the following technical effects:

The embodiment of the invention obtains the relationship between the throughput changes of the base stations according to the historical data of the base station, and constructs The base station relationship network selects a few important base stations from the network to evaluate the throughput of other large base stations. It has high practical value. For example, in the data acquisition of the base station, there are many base stations whose data is missing. With the embodiment of the present invention, the missing data can be evaluated for further network analysis. At the same time, it can be flexibly selected according to the historical data of the throughput of different regions or time periods to evaluate, with universal applicability and better prediction accuracy.

Although the embodiments of the present invention have been described in detail above, the present invention is not limited thereto, and various modifications may be made by those skilled in the art in accordance with the principles of the present invention. Therefore, modifications made in accordance with the principles of the invention are to be understood as falling within the scope of the invention.

One of ordinary skill in the art will appreciate that all or a portion of the above steps may be performed by a program to instruct related hardware, such as a processor, which may be stored in a computer readable storage medium, such as a read only memory, disk or optical disk. Wait. Alternatively, all or part of the steps of the above embodiments may also be implemented using one or more integrated circuits. Correspondingly, each module/unit in the foregoing embodiment may be implemented in the form of hardware, for example, by implementing an integrated circuit to implement its corresponding function, or may be implemented in the form of a software function module, for example, being executed by a processor and stored in a memory. Programs/instructions to implement their respective functions. The invention is not limited to any specific form of combination of hardware and software.

The embodiments disclosed in the present application are as described above, but the descriptions are only for the purpose of understanding the present application, and are not intended to limit the present application, such as the specific implementation method in the embodiments of the present invention. Any modifications and changes in the form and details of the embodiments may be made by those skilled in the art without departing from the spirit and scope of the disclosure. The scope defined by the appended claims shall prevail.

Industrial applicability

The above technical solution reduces the complexity of performing data.

Claims

A method for evaluating wireless network throughput, the evaluation method comprising:

Obtaining historical data of throughput of N base stations;

Constructing a base station relationship network of the N base stations according to the acquired historical data of the throughput of the N base stations;

Obtaining M base stations that play an important role in evaluating the throughput of the base station relationship network according to the constructed base station relationship network and the historical data of the obtained throughput, and using the M base stations as important base stations;

Using the historical data of the determined throughput of the M important base stations, evaluating the remaining N-M base station throughputs;

Where N and M are both positive integers and N is greater than M.
The evaluation method according to claim 1, wherein the obtaining historical data of the throughput of the N base stations comprises:

Collecting an original throughput sequence of each base station, and calculating an average value of the throughput sequence values in the collected original throughput sequence;

The value of the sequence of the preset percentage thresholds in which the sequence value is prior to each other in the original throughput sequence of each base station is replaced by the calculated average of the throughput sequence values. Value, get the new throughput sequence for each base station;

The normalized throughput sequence of each base station is obtained by normalizing the time series of the new throughput of each base station, and the obtained normalized throughput sequence is used as the historical data of the throughput.
The evaluation method according to claim 2, wherein the constructing the base station relationship network of the N base stations according to the historical data of the acquired throughput of the N base stations comprises:

Calculating correlation coefficients between two of the N base stations according to the normalized throughput sequence of each of the obtained base stations;

When the calculated correlation coefficient is greater than a correlation coefficient threshold, an undirected edge is generated between the two base stations;

Constructing the base station relationship network of the N base stations by using an undirected edge generated between two base stations of the N base stations.
The evaluation method according to claim 3, wherein the M base stations that use the constructed base station relationship network and the acquired historical data of the throughput to find an effect on the base station relationship network throughput evaluation effect include :

Obtaining the degree of each base station by counting the number of undirected edges of each base station in the base station relationship network;

Selecting the base stations of the N most basic base stations in sequence, and evaluating the throughput of the N-m base stations according to the support vector machine, and obtaining the throughput evaluation effect of the m base station relationship networks;

In the throughput evaluation effect of the obtained m base station relationship networks, the best throughput evaluation effect is selected, and the m base stations corresponding to the selected best throughput evaluation effects are used as the base station relationship network throughput evaluation. M base stations whose effects play an important role;

Where m, M, and N are positive integers, M<=m, M<N, and m<N.
The evaluation method according to claim 4, wherein the number of undirected edges of each base station is proportional to the size of the base station.
The evaluation method according to claim 5, wherein the evaluating the remaining N-M base station throughputs using the determined historical data of the throughput of the M important base stations comprises:

Constructing a throughput relationship model between the remaining N-M base stations and M important base stations by using a support vector machine algorithm;

The estimated throughput of the remaining N-M base stations is obtained by using the constructed throughput relationship model and the throughput history data of the M important base stations.
An apparatus for evaluating wireless network throughput, the evaluation apparatus comprising:

An acquisition module, configured to acquire historical data of throughput of N base stations;

a building module, configured to construct a base station relationship network of the N base stations according to the acquired historical data of the throughput of the N base stations;

a searching module, configured to use the constructed base station relationship network and the acquired historical data of the throughput to find M base stations that play an important role in evaluating the throughput of the base station relationship network, and use the M base stations as important base stations ;

An evaluation module, configured to use the historical data of the determined throughput of the M important base stations to evaluate the remaining N-M base station throughputs;

Where N and M are both positive integers and N is greater than M.
The evaluation device of claim 7, wherein the acquisition module comprises:

Calculating a throughput average unit, configured to collect a raw throughput sequence of each base station, and calculate an average value of the throughput sequence values in the collected original throughput sequence;

The obtaining unit is configured to replace the calculated throughput with a sequence of a preset percentage threshold of the sequence value before each sequence of the original throughput sequence of each of the acquired base stations is replaced by the calculated The average value of the sequence values is obtained, and a new throughput sequence of each base station is obtained, and the normalized throughput sequence of each base station is obtained by normalizing the time series of the new throughput of each base station, and the obtained The normalized throughput sequence is used as historical data for the throughput.
The evaluation device of claim 8, wherein the building block comprises:

Calculating the correlation coefficient unit is configured to calculate correlation coefficients between two of the N base stations according to the normalized throughput sequence of each of the obtained base stations;

Generating the undirected edge unit to set an undirected edge between the two base stations when the calculated correlation coefficient is greater than the correlation coefficient threshold;

The building unit is configured to construct the base station relationship network of the N base stations by using an undirected edge generated between two base stations of the N base stations.
The evaluation device of claim 9, wherein the lookup module comprises:

The obtaining unit is configured to obtain the degree of each base station by counting the number of undirected edges of each base station in the base station relationship network, and sequentially select the m base stations with the medium most middle of the N base stations, according to the support vector machine Assessing the throughput of the Nm base stations, and obtaining the throughput evaluation effect of the m base station relationship networks;

The searching unit is configured to select the best throughput evaluation effect in the throughput evaluation effect of the obtained m base station relationship networks, and use the m base stations corresponding to the selected best throughput evaluation effect as the base station M base stations that play an important role in the network throughput evaluation effect;

Where m, M, and N are positive integers, M<=m, M<N, and m<N.