CN113486301B - Method for extracting network ecological factors - Google Patents

Abstract

The invention belongs to the technical field of communication, and particularly relates to a method for extracting a network ecological factor. The method mainly comprises the steps of firstly obtaining a network factor set X of factors influencing the environment, network elements, users and the like of a future network architecture, and then constructing a base network factor model by adopting random vectors which cannot be observed, namely network common factors; calculating sample mean from network factor data array

A sample covariance matrix S and a sample correlation matrix R; solving the eigenvalue and the standardized eigenvector of a sample correlation matrix R of the network factor; solving a network factor load matrix A of the network factor model; and (4) carrying out rotation transformation on the network factor load matrix by using a factor rotation method to obtain a rotated network factor load matrix, and finally extracting the network ecological factors. The invention has the beneficial effects that: the amount of data that the network needs to process is greatly reduced.

Description

Method for extracting network ecological factors

Technical Field

The invention belongs to the technical field of communication, and particularly relates to a method for extracting a network ecological factor.

Background

The wireless environment is a very important factor affecting mobile communication networks. The variety of wireless environments is also very complex. First, consider the location of a web site, such as whether the site is located in the air, sky, ground, or sea, whether it is a city or countryside, whether it is a mountain, plain, forest, or sea; secondly, the speed of the mobile terminal is high speed, medium speed or low speed, etc.; finally, what the data related to the user is, such as whether the user model is dense or sparse, user preference tendencies, and the like.

No matter how the architecture of the mobile communication network changes, the basic factors which are separated to influence the network architecture are the environmental factors, the network element factors, the user factors and the like. The environment factors correspond to models in different classification modes due to different classification modes, such as an urban environment model and a rural environment model, a mountain environment model and a forest environment model, a marine environment model and an aerospace environment model and the like; the network element factors relate to network element capacity, network element speed and the like, wherein the network element capacity refers to the processing capacity of network equipment or a terminal; user factors are primarily related to user models such as user demand, user density, user preferences, user speed, etc.

These factors are only qualitative indications. Nowadays, the comprehensive application of artificial intelligence technology in future networks is well known in the industry, but for real implementation, the factors are required to be mathematically modeled and converted into data which can be identified and processed by a computer. All variables involved in the mathematical modeling process of possible factors affecting the future network structure are defined as network factors.

The aforementioned differences of the wireless environments bring differences of wireless environment propagation models, but the differences are only one of factors influencing the design of future network structures. In fact, there are also a large number of influencing factors in future networks, such as network elements, users, etc. By further analyzing a plurality of elements related to the factors, more network factors influencing the setting of the network station and the configuration of the station parameters can be obtained, so that a large number of network factors influencing the future network topology design such as the selection of the type, the position and the number of the stations and the configuration of the future network parameters (such as the working frequency of the stations, the channel configuration parameters, the channel allocation parameters, the switching conditions, the antenna configuration and the like) can be obtained.

However, there may be some associations between the network factors, so that their effects on the network are not independent of each other. And the factors which are independent from each other or really influence the network independently are screened from the possible network factors through certain theoretical modeling, and are defined as the network ecological factors.

Disclosure of Invention

Aiming at the problems, the invention provides a method for extracting a network ecological factor from a network factor, which is how to extract the network ecological factor from a plurality of network factors.

The technical scheme of the invention is as follows:

(1) firstly, obtaining a network factor set X of factors influencing the environment, network elements, users and the like of a future network architecture:

X＝{X₁,X₂,...,X_p}′

wherein X is a random vector of observable network factors, wherein X is_pThe network factor is the p-th network factor, and the network element factor, the user factor and the like are indicated;

then, adopting random vectors which cannot be observed, namely network public factors, to construct a base network factor model;

(2) calculating sample mean from network factor data array

A sample covariance matrix S and a sample correlation matrix R;

(3) obtaining a factor load matrix A of the network factor model according to the correlation matrix R;

(4) and extracting the network ecological factors.

The invention has the beneficial effects that: the amount of data that the network needs to process is greatly reduced.

Drawings

Fig. 1 is an example of factors affecting a network.

Fig. 2 is a schematic diagram of the extraction from the network factors to the network ecological factors.

Detailed Description

The present invention will be described in detail with reference to examples.

Examples

As shown in fig. 1, in the method of this embodiment, a network factor set X that affects factors such as environment, network elements, and users of a future network architecture is obtained first: x ═ X₁,X₂,...,X_p}′

Wherein X is a random vector of observable network factors, wherein X is_pThe p-th network factor refers to environment factor, network element factor, user factor, etc., and the mean value is e (x) ═ μ₁,μ₂,...,μ_p) Assistant prescriptionThe difference matrix is d (x) ═ Σ.

Let F ═ F₁,F₂,...,F_m)′(m<p) is an unobservable random vector, E (F) 0, D (F) I_mIf epsilon is equal to (epsilon)₁,ε₂,...,ε_p) ' is independent of F, and E (ε) ═ 0,

assume that network factor X can be expressed as:

the above formula can be expressed as a matrix:

X＝μ+AF+ε (2)

wherein F ═ F₁,F₂,...,F_m) ' is a common factor of the network factor X, e ═ e [ (. epsilon. ])₁,ε₂,...,ε_p) ' is a special factor of the network factor X, A ═ a_ij)_p×mFor a coefficient matrix to be estimated, called network factor load matrix, a_ijThe load of the ith variable on the jth network factor is referred to as (i 1.. cndot.p.; j 1.. cndot.m).

(II) calculating sample mean value by network factor data array

A sample covariance matrix S and a sample correlation matrix R.

Let the sample data array be

Then the sample mean is

Sample covariance matrix of

Wherein

The sample correlation matrix is

R＝(r_ij) (6)

Wherein

(III) solving a factor load matrix A of the network factor model according to the sample correlation matrix R

(IV) extracting the network ecological factors, as shown in figure 2:

in the formula (2), it is assumed that

Following an (m + p) metanormal distribution, then:

it also follows a (m + p) normal distribution, from which the mean and covariance matrices can be derived as:

under the conditions given by X, the conditional mathematical expectation for F is

(10) Can be represented as

Or

By using

R is used to obtain a factor score instead of μ, A, Σ in the formula (10), and X is therefore used_jIs scored as

(13) In (1)

Namely the network ecological factor.

Claims (1)

1. A method for extracting network ecological factors is characterized by comprising the following steps:

s1, acquiring a network factor set X of the environment, the network elements and the users influencing the network architecture:

X＝{X₁,X₂,...,X_p}′

wherein X is a random vector of observable network factors, wherein X is_pThe p-th network factor has the mean value of E (X) mu (mu)₁,μ₂,...,μ_p) The covariance matrix is d (x) ═ Σ;

if F ═ F₁,F₂,...,F_m) ' is a random vector that is not observable, m<p，E(F)＝0，D(F)＝I_mAnd a special factor of epsilon ═ epsilon₁,ε₂,...,ε_p) Is independent of F, and

σ²for the variance of a particular factor, the network factor X is expressed as:

the above formula can be expressed as a matrix:

X＝μ+AF+ε

wherein A ═ a_ij)_p×mFor a coefficient matrix to be estimated, called network factor load matrix, a_ijThe load of the ith variable on the jth network factor, i 1., p; j 1, a, m;

s2, calculating sample mean value by network factor data array

Sample covariance matrix S and sample correlation matrix R:

let the sample data array be

Where n is the number of samples, the mean of the samples is

The sample covariance matrix is

Wherein

The sample correlation matrix is

R＝(r_ij)

Wherein

S3, obtaining a factor load matrix A of the network factor model according to the correlation matrix R;

s4, suppose

Following an (m + p) metanormal distribution, then:

it also follows a (m + p) normal distribution, from which the mean and covariance matrices can be derived as:

under the conditions given by X, the conditional mathematical expectation for F is

Is shown as

Or

By using

R replaces μ, A, Σ in the formula to find a factor score, so X_jIs scored as

Obtained by

Namely the network ecological factor.

Images