CN115243250B

CN115243250B - Method, system and storage medium for acquiring wifi portrait

Info

Publication number: CN115243250B
Application number: CN202210880497.5A
Authority: CN
Inventors: 尹雅露; 莫志强; 陈志勇; 方宏源
Original assignee: Merit Interactive Co Ltd
Current assignee: Merit Interactive Co Ltd
Priority date: 2022-07-25
Filing date: 2022-07-25
Publication date: 2024-05-28
Anticipated expiration: 2042-07-25
Also published as: CN115243250A

Abstract

The application provides a method, a system and a storage medium for acquiring wifi portraits, which are characterized in that firstly, the SSID name of wifi containing Chinese characters is acquired, then the SSID name is input into a preset wifi classification model to obtain the category portraits of the wifi, wherein the preset wifi classification model is obtained by training a plurality of different text-cnn models based on a plurality of wifi subsets for equivalent training which are randomly extracted from a training wifi set and a category subset to which the wifi subsets belong. Through the content, the robustness of the preset wifi classification model can be obviously improved, so that the acquired wifi class portrait is more accurate.

Description

Method, system and storage medium for acquiring wifi portrait

Technical Field

The application relates to the field of data processing, in particular to a method, a system and a storage medium for acquiring wifi portraits.

Background

Wifi plays an increasingly important role in people's daily work and life due to its own convenience. Because wifi self position is usually more fixed, therefore, the user portrait of describing wifi portrait helps further to help the construction connect this wifi, wherein acquire wifi portrait through wifi naming (SSID name) is a common method. In the prior art, people are used to name wifi by using Chinese characters and other characters, so that some wifi can be named according to Chinese characters easily to obtain a category portrait of the wifi, for example, the wifi of a market class can be named as a certain market, but even if the Chinese characters are used, a considerable part of wifi is still named, the category attribute of the wifi is still difficult to identify according to the naming of the wifi, and therefore, how to obtain the wifi portrait named by the Chinese characters is a technical problem to be solved at present.

Disclosure of Invention

Aiming at the technical problems, the application adopts the following technical scheme: a method for acquiring wifi portrait comprises the following steps: s100, acquiring an SSID name of the wifi, wherein the SSID name contains Chinese characters; s200, obtaining the class portrait of the wifi based on the SSID name and a preset wifi classification model.

A system for capturing wifi portraits, the system comprising a processor and a non-transitory computer readable storage medium storing at least one instruction or at least one program, the processor loading and executing the at least one instruction or at least one program to implement the method described above.

A computer-readable storage medium storing a program or instructions that cause a computer to perform the steps of the method described above.

According to the method, firstly, the SSID name of the wifi containing Chinese characters is obtained, and then the SSID name is input into a preset wifi classification model to obtain the class portrait of the wifi. The preset wifi classification model is obtained by training a plurality of different text-cnn models based on a plurality of wifi subsets for equivalent training, which are randomly extracted from a wifi set for training, and a category subset to which the wifi subsets belong. Through the content, the robustness of the preset wifi classification model can be improved, so that the acquired wifi class portrait is more accurate.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flowchart of a method for obtaining wifi portrait according to an embodiment of the present application.

Detailed Description

The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to fall within the scope of the application.

The embodiment of the application provides a method for acquiring wifi portrait, which is shown in figure 1 and comprises the following steps:

S100, acquiring the SSID name of the wifi, wherein the SSID name contains Chinese characters. Those skilled in the art can use the SSID as the wireless network name of wifi, and the user can change the SSID name according to the need. In this step, a regular manner may be used to determine whether the SSID name contains chinese characters, and whether the SSID name contains chinese characters may also be determined according to a preset chinese character library, where in one embodiment, the preset chinese character library includes all chinese characters, and in another embodiment, the preset chinese character library may be determined according to chinese characters contained in the SSID names of all wifi acquired.

S200, obtaining the class portrait of the wifi based on the SSID name and a preset wifi classification model. In the present application, any classification technology in the prior art, such as SVM, may be used as the preset wifi classification model, and in the preferred embodiment of the present application, the preset wifi classification model is a text-cnn model. Specifically, the obtaining of the preset wifi classification model includes:

S301, obtaining an SSID name set S= { SSID ₁,SSID₂,...,SSID_n } of a training wifi set and a corresponding belonging category set L= { L ₁,L₂,...,L_n }, wherein the SSID name of the ith wifi in the training wifi set is SSID _i, the corresponding belonging category is L _i,L_i and is one of M preset categories, and SSID _i contains Chinese characters, and i is more than or equal to 1 and less than or equal to n. In the present application, M preset categories may be set in a customized manner, for example, traffic category, peripheral tour category, hotel category, etc., and the above is only exemplary, and is not intended to limit the scope of the present application. Those skilled in the art will recognize that, in order to improve the robustness of the classification model, the training wifi sets as many samples of each of the M preset categories as possible, so as to traverse all possible situations.

S302, training the text-cnn model by taking the name set S and the belonging class set L as input data of the text-cnn to obtain the preset wifi classification model. Specifically, the method comprises the following steps:

s3021, obtaining an orientation quantized name set sv= { SSIDV ₁,SSIDV₂,...,SSIDV_n } based on the name set S. Specifically, in the present application, any vectorization method in the prior art may be used to vectorize SSID _i to SSIDV _i, such as one-hot, word2vec, and so on. In this step, the person skilled in the art can count all Chinese characters and non-Chinese characters of the SSID name that may be used for wifi and vectorize the SSID name based on the counted total number of characters. For example, when the number of all the counted characters (assuming that only 5 characters in north and south of east) is 5 and the SSID names are vectorized using the one-hot method, vectorized data corresponding to the 5 characters (in north and south of west) may be set to 10000, 01000, 00100, 00010, 00001, and when the SSID name of a wifi is "southeast", the data obtained by vectorizing the SSID name of the wifi is a2×5 matrix: [ [1, 0]; [0,0,1,0,0]].

S3022, an equal length name set sh= { SSIDH ₁,SSIDH₂,...,SSIDH_n } is obtained based on the vectorized name set sv= { SSIDV ₁,SSIDV₂,...,SSIDV_n }. In this step, the maximum length of the element in the wifi name set may be obtained according to the preset maximum length of the wifi name set or based on the vectorized name set SV, and the 0-compensating operation is adopted so that SSIDH ₁,SSIDH₂,...,SSIDH_n has the same matrix line number. For example, according to the example in step S3021, if the maximum length of the element in the SV or the maximum length of the preset wifi name is 4, the SSID name "southeast" becomes [ [1, 0] after the 0-supplementing operation; [0, 1, 0]; [0, 0]; [0, 0] ], i.e., a 4×5 matrix.

S3023, inputting the equal-length name set SH= { SSIDH ₁,SSIDH₂,...,SSIDH_n } and the category set L into a text-cnn model for training to obtain the preset wifi classification model. In order to obtain the optimal wifi category portrait, the convolution kernel size of the text-cnn model has a value range of [ three, five ], preferably five, and 128 convolution kernels each. Preferably, the convolution kernels are convolution kernels with dimensions 1, 2, 3,4, 5, respectively. By using the five convolution kernels with different sizes, the effective characteristics in the SSID names of each wifi can be comprehensively and effectively obtained, and the classification accuracy is further improved. Table 1 shows the partial recognition accuracy data of images of different wifi categories according to the application.

TABLE 1

	Predictive score >0	Predictive score >0.5	Predictive score >0.6	Predictive score >0.7
					Accuracy of	53％	66％-70％	73.5％-75％	75％-78％
Coverage rate	100％	80％	65％	50％

As can be seen from table 1, for a given test set, relative prediction accuracy can be obtained according to different prediction score thresholds, when the prediction score threshold=0, the prediction score of 100% of samples in the given test set is > 0, and the prediction accuracy of the portion of samples with the sample prediction score > 0 is 53%; when the predictive value threshold = 0.5, the predictive value of 80% of the samples in the given test set is > 0.5, and the predictive accuracy of the portion of samples with the predictive value of > 0.5 is 66% -70% (this predictive accuracy is due to some wifi that cannot be objectively verified, which leads to the existence of the interval, that is, if the wifi that cannot be objectively verified is considered to be correctly identified, the predictive accuracy at this time is 70%, otherwise 66%); when the predictive value threshold = 0.6, the predictive value of 65% of the samples in the given test set is > 0.6, and the predictive accuracy of the portion of samples having a sample predictive value > 0.6 is 73.5% -75%; when the predictive value threshold=0.7, the predictive value of 50% of the samples in the given test set is > 0.7, and the predictive accuracy of the portion of samples having a sample predictive value > 0.7 is 50%. According to the method, firstly, the SSID name of the wifi containing Chinese characters is obtained, and then the SSID name is input into a preset wifi classification model to obtain the class portrait of the wifi. Through the preset wifi classification model, the class portrait of wifi can be effectively obtained.

In a preferred embodiment of the present application, the obtaining the preset wifi classification model includes:

S401, obtaining an SSID name set S= { SSID ₁,SSID₂,...,SSID_n } of a training wifi set and a corresponding belonging category set L= { L ₁,L₂,...,L_n }, wherein the SSID name of the ith wifi in the training wifi set is SSID _i, the corresponding belonging category is L _i,L_i and is one of M preset categories, and SSID _i contains Chinese characters, and i is more than or equal to 1 and less than or equal to n.

S402, acquiring SSID name subsets S ¹、S²、...、S^K of K groups of wifi subsets for training and corresponding class subsets L ¹、L²、...、L^K based on the SSID name set S and the belonging class set L, wherein m obtained name subsets S ^j＝{SSID^j ₁,SSID^j ₂,...,SSID^j _m are randomly extracted from the SSID name set S, and the corresponding belonging class of the corresponding belonging class subset L ^j＝{L^j ₁,L^j ₂,...,L^j _m},SSID^j _t is L ^j _t,L^j _t E L, m/n=a preset grouping ratio threshold value, j is more than or equal to 1 and less than or equal to K, and t is more than or equal to 1 and less than or equal to m. Specifically, in the present application, the preset packet duty ratio threshold has a value range of [0.7,0.9], preferably 0.8.K has a value of [3,5], preferably 5.

S403, respectively inputting the SSID name subset S ¹、S²、...、S^K and the corresponding category subset L ¹、L²、...、L^K into K different text-cnn classification models for training to obtain the preset wifi classification model. In the step, S ¹ and L ¹ are input into a1 st text-cnn classification model for training to obtain a1 st trained text-cnn model, S ² and L ² are input into a2 nd text-cnn classification model for training to obtain a2 nd trained text-cnn model, and so on, so as to obtain K different trained text-cnn models. When a class portrait of wifi needs to be predicted, SSID names of the wifi are respectively input into the K trained different text-cnn models to obtain K different class predicted values, voting is conducted based on the K different class predicted values, and finally the class portrait of wifi is obtained. The voting mechanism may employ any of the prior art techniques, such as the commonly used minority-compliance majority.

In order to obtain the optimal wifi category portrait, the convolution kernel size of the text-cnn model has a value range of [ three, five ], preferably five, 128 of each. Preferably, the convolution kernels are convolution kernels with dimensions 1, 2, 3,4, 5, respectively. By using the five convolution kernels with different sizes, the effective characteristics in the SSID names of each wifi can be comprehensively and effectively obtained, and the classification accuracy is further improved. Table 2 is the partial recognition accuracy data of the images of different wifi categories based on the obtained preset wifi classification model.

TABLE 2

	Predictive score >0	Predictive score >0.5	Predictive score >0.6	Predictive score >0.7
					Accuracy of	58.9％-72.1％	69.6％-78.5％	72.8％-79.4％	78.3％-83.9％
Coverage rate	100％	86％	73％	59％

According to the contents of table 2, the predicted result of the preset wifi classification model is obviously better than the data in table 1 in terms of coverage rate and accuracy, and the effect of the preset wifi classification model obtained by the above preferred embodiment is better. Table 3 shows partial wifi prediction accuracy data of the preset wifi classification model according to the preferred embodiment of the present application for different wifi class portraits.

TABLE 3 Table 3

Category portrayal	Classification prediction accuracy
		Individuals	0.9
Leisure and recreation	0.9
		Company enterprise	0.8
Peripheral game	0.6
		Colleges and universities	0.9
Household building material	0.7
		Food for delicacies	0.6
Shopping	0.8
		......	......

From the content of Table 3, the method provided by the application can accurately acquire the wifi class portrait, can meet the requirements of people on the wifi class portrait to a certain extent, and has strong adaptability.

The embodiment of the application also provides a system for acquiring wifi portrait, which comprises a processor and a non-transitory computer readable storage medium, wherein the storage medium is used for storing at least one instruction or at least one section of program, and the processor loads and executes the at least one instruction or the at least one section of program to realize the method provided by the embodiment.

A computer-readable storage medium storing a program or instructions that cause a computer to execute the method provided by the above-described embodiment.

Embodiments of the present application also provide a non-transitory computer readable storage medium that may be disposed in an electronic device to store at least one instruction or at least one program for implementing one of the methods embodiments, the at least one instruction or the at least one program being loaded and executed by the processor to implement the methods provided by the embodiments described above.

Embodiments of the present application also provide an electronic device comprising a processor and the aforementioned non-transitory computer-readable storage medium.

Embodiments of the present application also provide a computer program product comprising program code for causing an electronic device to carry out the steps of the method according to the various exemplary embodiments of the application as described in the specification, when said program product is run on the electronic device.

While certain specific embodiments of the application have been described in detail by way of example, it will be appreciated by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the application. Those skilled in the art will also appreciate that many modifications may be made to the embodiments without departing from the scope and spirit of the application. The scope of the application is defined by the appended claims.

Claims

1. The method for acquiring the wifi image is characterized by comprising the following steps of:

s100, acquiring an SSID name of the wifi, wherein the SSID name contains Chinese characters;

S200, acquiring the class portraits of the wifi based on the SSID name and a preset wifi classification model;

the obtaining of the preset wifi classification model comprises the following steps:

S401, acquiring an SSID name set S= { SSID ₁,SSID₂,...,SSID_n } of a training wifi set and a corresponding belonging category set L= { L ₁,L₂,...,L_n }, wherein the SSID name of the ith wifi in the training wifi set is SSID _i, the corresponding belonging category is L _i,L_i and is one of M preset categories, and SSID _i contains Chinese characters, and i is more than or equal to 1 and less than or equal to n;

S402, acquiring SSID name subsets S ¹、S²、...、S^K of K groups of wifi subsets for training and corresponding class subsets L ¹、L²、...、L^K based on the SSID name set S and the belonging class set L, wherein m obtained name subsets S ^j＝{SSID^j ₁,SSID^j ₂,...,SSID^j _m are randomly extracted from the SSID name set S, the corresponding belonging class of the corresponding belonging class subset L ^j＝{L^j ₁,L^j ₂,...,L^j _m},SSID^j _t is L ^j _t,L^j _t E L, m/n=a preset grouping ratio threshold value, j is more than or equal to 1 and less than or equal to K, and t is more than or equal to 1 and less than or equal to m;

S403, respectively inputting the SSID name subset S ¹、S²、...、S^K and the corresponding category subset L ¹、L²、...、L^K into K different text-cnn classification models for training to obtain the preset wifi classification model; further, when predicting a class portrait of wifi, respectively inputting SSID names of the wifi into K trained different text-cnn models to obtain K different class predicted values, voting based on the K different class predicted values, and finally obtaining the class portrait of wifi;

The value range of the preset grouping ratio threshold is [0.7,0.9], the convolution kernel size corresponding to the text-cnn model is five, and 128 convolution kernels are respectively arranged; the convolution kernels are convolution kernels with dimensions 1,2, 3, 4, 5, respectively.

2. The method of claim 1, wherein K has a value of [3,5].

3. A system for capturing wifi portraits, the system comprising a processor and a non-transitory computer readable storage medium storing at least one instruction or at least one program, wherein the processor loads and executes the at least one instruction or at least one program to implement the method of claim 1 or 2.

4. A computer readable storage medium storing a program or instructions for causing a computer to perform the steps of the method according to claim 1 or 2.