TWI556121B - Gender prediction method by using webpage surfing behavior - Google Patents

Description

Gender prediction method using web browsing behavior

The following narrative is about a method of gender prediction, and in particular, a method for predicting gender using a user's web browsing behavior on a blog website.

Blog sites usually refer to sites that are sorted by time and updated from time to time, most of which exist in a diary and contain articles and reverberation. The main axis of a blog can be a personal life, even a political topic, and its theme can be a niche issue or a wide range of topics.

In the current well-known blog sites, the advertising revenue on the webpage is mainly one of its sources of income, and if it is to attract well-known advertisers to join its website, the website must have a very high click rate or Popularity. However, whether it is the advertisement on the website or the content of the website inside it, most of them are currently played in one-way playback mode, that is, the file of the webpage is repeatedly played back in a certain period of time, but this way is very likely Inducing the user's resentment, and reducing the click rate. For example, when a male user browses a blog website and sees the content of the article or advertisement information about the woman, the male user may jump out of the blog site impatiently without continuing. Browse.

The main reason for the occurrence of this situation is that this website cannot be known in front of the computer. The gender of the user, and then play the associated advertisement or article according to their gender, so that the user can stay in the blog website. Because of this, some of the current blog sites may be biased towards a specific theme, such as only attracting a part of the teenage ethnic group or the working group, which cannot be widely applied to different users.

As a result, for blog sites, the source of their advertisers can only be restricted to one area, which will lead to a decrease in their income. For the user, it can only see the content of the article and the advertisements that are fixed by the blog site. Moreover, for advertisers, their marketing channels cannot be effectively expanded, resulting in a three-loss situation.

Therefore, the inventors have proposed a method of performing gender prediction through a blog website to solve the above problems.

In view of the above-mentioned problems, the present invention is to solve the problem that it is impossible to know the gender of a user browsing a blog website.

In view of the above-mentioned problems, the present invention is to solve the problem that the appropriate information and advertisement content cannot be played for the user.

Based on the above object, the present invention provides a gender prediction method using web browsing behavior, which is applicable to a blog website, which is installed on a web server, and the website server stores a plurality of first uses respectively. The gender information of the person and one of the web browsing information of the plurality of first users when logging into the blog website, the gender prediction method comprises the steps of: discretizing a plurality of consecutive features of the first user to form at least one discrete feature, the continuous feature may Contains a continuous value. Generating a plurality of test features according to the at least one discrete feature and the plurality of first features, the plurality of first features including an article category, a login source URL, an article browsing information, and a previous website information when logging into the blog site. Calculate the proportion of males and the proportion of females corresponding to a plurality of test characteristics. Multiple seconds when the second user browses the blog site The feature is substituted into the classifier, and the classifier calculates the sex ratio of one of the second users according to the plurality of male ratios and the plurality of female proportions. The plurality of first users and the plurality of second users continuously browse the webpage twice or more through the blog website.

Preferably, the gender prediction method of the present invention further comprises using a Laplace Smoothing to normalize the male ratio and the female proportion.

Preferably, the classifier may comprise a Naïve Bayes Classifier.

Preferably, the login source URL may include a country information or an organization information.

Preferably, the article browsing information may include an author information.

Preferably, the continuous feature can be one of the time to browse the blog website.

Preferably, the gender prediction method of the present invention further comprises using a hypothesis test method to reduce the number of the plurality of first features, and the hypothesis test method can be a chi-square test.

Preferably, the previous website information may include an external web page different from one of the blog websites and another web page within the blog website when logging into the blog website.

Preferably, the gender prediction method of the present invention further comprises an instant change of the content of the advertisement in the blog website, the design of the layout, and the recommended article according to the gender ratio.

100‧‧‧Blog Website

101‧‧‧Web server

20‧‧‧ second user

21‧‧‧ second feature

22‧‧‧Update page

S11~S14‧‧‧Steps

The above and other features and advantages of the present invention will become more apparent from the detailed description of the embodiments of the invention.

2 is a first schematic diagram of a gender prediction method according to another embodiment of the present invention.

Figure 3 is a second schematic diagram of a gender prediction method according to another embodiment of the present invention. Figure.

The features, the contents and advantages of the present invention, and the advantages thereof, will be understood by the present invention. The present invention will be described in detail with reference to the accompanying drawings, The use of the present invention is not intended to be a limitation of the scope of the present invention, and the scope of the present invention is not limited by the scope and configuration of the accompanying drawings.

The advantages and features of the present invention, as well as the technical methods of the present invention, are described in more detail with reference to the exemplary embodiments and the accompanying drawings, and the present invention may be implemented in various forms and should not be construed as limited thereby. The embodiments of the present invention, and the embodiments of the present invention are intended to provide a more complete and complete and complete disclosure of the scope of the present invention, and The scope of the patent application is defined.

Please refer to Fig. 1, which is a flow chart of a gender prediction method according to the present invention. As shown in the figure, the gender prediction method using web browsing behavior is applied to a blog website, which is installed on a web server, and the web server can be a workstation host or a computer host and its It is possible to store a plurality of first user gender data and one of the plurality of first users logging into the blog website, and the gender information can be recorded when the first user joins the member of the blog website, and After each time the first user logs into the blog website, the first user's browsing history, such as its IP location, the selected web page, and the like, are also stored when the user browses the web page. The gender prediction method of the present invention comprises the following steps.

Step S11 discretizes one of the plurality of first users to form at least one discrete feature, wherein the continuous feature may include a continuous value, such as the time of login or the sum of time of browsing the webpage.

Step S12: generating a plurality of test features according to the at least one discrete feature and the plurality of first features, wherein the plurality of first features may include an article category, a login source URL, an article browsing information, and a previous login to the blog site Website information, the login source URL can include a country information (such as .tw or cn) or an organization information (such as com or org). The article browsing information can include an author information. When logging in to the blog website, the previous website information can include An external webpage that is different from one of the blog sites, that is, an external webpage links to the blog site, and another webpage within the blog site, that is, another webpage of the blog site is linked to the currently viewed webpage.

Step S13 is to calculate a plurality of male proportions and a plurality of female proportions corresponding to the plurality of test characteristics, and substitute a plurality of male proportions and a plurality of female proportions into a classifier.

Step S14 is to substitute a plurality of second features when a second user browses the blog website into a classifier, and the classifier calculates a gender ratio of the second user according to the plurality of male ratios and the plurality of female proportions. The classifier can be a Naïve Bayes Classifier, and the plurality of second features are included in the plurality of test features, and the plurality of first users and the plurality of second users are transmitted through The blog site continuously browses the web page twice or more.

The gender prediction method of the present invention may further include the gender ratio calculated according to step 14 to instantly change the content of the advertisement in the blog website, the design of the layout, and the recommended article.

The gender prediction method of the present invention may further comprise using a hypothesis testing method to reduce the number of the plurality of first features, wherein the hypothesis testing method may be a chi-square test.

As can be seen from the above, the gender prediction method of the present invention can predict the non-member usage of browsing the blog website through the web browsing behavior of the members registered on the blog website. The gender of the person, and immediately generate content, advertisements, etc. related to the user's gender based on the predicted gender, to specifically improve the problem that the traditional blog website cannot play appropriate information and advertisement content for the user.

Another preferred embodiment of the present invention is embodied by a login member and a non-login member of a blog site, wherein the blog site is erected on a web server. This website server records the browsing information of the member when logging into the blog website, including the member's name, gender, age, nature of work, age and country, the ip address of the login, the browser used, and the tribe in the tribe. The pages viewed by the website, the blog posts and their categories, the login time, the blog page before and after browsing a web page, and so on.

Table 1 shows the categories and login times of the five members (p1~p5) recorded when logging into this blog site. As shown in Table 1, p1 and p2 are male users, while p3~p5 are female. user. It is worth mentioning that the content of Table 1 is the result of the discretization record. The discretization method is based on cutting the login time of one day into 24 equal parts, that is, distinguishing by hour, and recording the member every hour. For the category of articles viewed, for example, male users of p1 have browsed articles on the "finance" and "sports" categories in the time zone of 06:00AM and 07:00AM, while the female users of p3 were at 03:00PM. And at 04:00PM, we will browse the articles on "Makeup" and "Makeup" and "Travel" respectively. However, the "Browse Article Category" and "Login Time" selected in Table 1 are only examples, but they are not limited to this. They can also be implemented by the characteristics of other users' browsing behaviors, such as the blog posts viewed. Author and category, a web page information or country before browsing a web page (ie, presumping with ip).

Table 2 shows the characteristic characteristics table compiled by Table 1. Among the characteristics of the feature "hour_6", the male characteristics "2/2" means "in 2 male members, both male members have Login at 6:00, and in the female feature "1/3" means "only one of the 3 female members will log in at 6:00", this calculation method is to calculate the "Browse Article Category" and The "login time" relative to the "gender" conditional probability value, the other characteristics of the corresponding male and female characteristics are the same as the feature "hour_6" and its corresponding male and female characteristics, so here Do not repeat them.

Further, Table 2 can be subjected to Laplace Smoothing to normalize the male ratio and the female proportion to form Table 3 below. This Laplacian smooth normalization method utilizes the numerator plus 1 and the denominator plus 2 ways to perform normalization operations.

After establishing the male ratio and the female ratio in Table 3, the web server can store the result in the storage unit therein, wherein the storage unit can be a physical memory or a hard disk. When a non-member user browses the blog site, the web server can use the male ratio and the female ratio stored in Table 3 to predict the gender of the non-member user.

For example, when a non-member user p6 browses the blog website, if the user browses the article category of sports and travel at 06:30AM and 07:15AM, then the user p6 is recorded as "browsing". The article category and login time can be first discretized as shown in Table 4.

In the present embodiment, a Naïve Bayes Classifier is used as a classifier, and the formula is as follows: P ( x | y _j ) = Π _{i = n} ( P ( i | y _j ) × x _i +(1- x _i )(1- P ( i | y _i ))), i {1,2,~ n }

Among them, in order to facilitate the calculation, we take the above formula as a logarithm to calculate, and according to the above two formulas, we can calculate the user p6 as a male score of -2.3 (ie log (2/5) + log ( 3/4)+log(2/4)+log(1-2/4)+log(1-1/4)+log(1-1/4)+log(1-1/4)+log( 1-2/4)+log(3/4)+log(2/4)+log(1-1/4)~=-2.3), where the first one (2/5) represents five users The probability of being a male, and taking the values corresponding to six points, seven points, sports and tourism, and 3/4, 2/4, 3/4 and 2/4 respectively, the other values are all subtracted by 1 (1-2/4), (1-1/4), (1-1/4), (1-1/4), (1-2/4), and (1-1/4), for women The score is -3.54 (log(3/5)+log(2/5)+log(1/5)+log(1-1/5)+log(1-3/5)+log(1-2 /5) +log(1-2/5)+log(1-1/5)+log(2/5)+log(2/5)+log(1-3/5)~=-3.54), The first one (3/5) indicates the probability of being a female among the five users, and takes the values corresponding to six points, seven points, sports and travel, and is 2/5, 1/5, 2/ respectively. 5 and 2/5, other values are always subtracted by 1 (1-1/5), (1-3/5), (1-2/5), (1-2/5), (1 -1/5) and (1-3/5), and because of the male score , Which is transmitted through sex predictive methods of the invention will be predictable user p6 The gender is a male.

After the web server calculates that the user p6 is a male, it can update the webpage that the user p6 is browsing, such as updating the advertisement content of the male user, including sports equipment, 3C supplies, or a car. Advertising, etc., or updated to a blog post that is of interest to male users, to increase the browsing time of p6 staying on the blog site, thereby increasing the click rate or popularity of the blog site.

It is worth mentioning that the web browsing record of the user p1~p6 used in the present invention means that the webpage must be browsed twice or more on the blog website. In other words, the user must stay on the blog website. And at least one page switching action is performed. If the user clicks directly on the blog site and then leaves the blog site directly, such web browsing history will not be used in the present invention.

Please refer to FIG. 2 and FIG. 3 , which are a first schematic diagram and a second schematic diagram of a gender prediction method according to another embodiment of the present invention. In this embodiment, the blog website 100 is installed on a website server 101. The website server 101 can include a computer host, a workstation, or a server. As shown in FIG. 2, when a second user 20 browses to the webpage of the blog website 100 of the present invention by using a computer, if the second user 20 does not log in to the blog website 100 as a member. And when the website server 101 browses at least two or more pages, the website server 101 can receive the packet sent by the second user 20 when the webpage is clicked through the network to learn the second user 20 A plurality of second features 21 when browsing the webpage, wherein the second feature 21 may include a time of login, an article category, a login source URL, a category of the article being viewed, an author, and other related information.

As shown in FIG. 3, after the website server 101 receives the second feature 21, the gender prediction method of the present invention can be used to predict the gender of the second user 20 who is currently browsing the blog website 100. And it utilizes the network of the first user The calculation method of the page browsing information for prediction is disclosed in the above embodiment, and therefore will not be described herein. When it is predicted that the gender of the second user 20 is female, then the blog website 100 can transmit an update page 22, including a female advertisement page, a blog article page, or update to a more flexible one. The layout is designed to be viewed by the second user 20 to increase the dwell time or click rate of the second user 20 on the blog website 100.

The embodiments described above are merely illustrative of the technical spirit and the features of the present invention, and the objects of the present invention can be understood by those skilled in the art, and the scope of the present invention cannot be limited thereto. That is, the equivalent variations or modifications made by the spirit of the present invention should still be included in the scope of the present invention.

S11~S14‧‧‧ Process steps

Claims (8)

A gender prediction method using web browsing behavior is applicable to a blog website, the blog website is installed on a web server, and the website server stores a plurality of first user gender data and the plurality of web services respectively. a web browsing information when the first user logs into the blog website, the gender prediction method includes: discretizing one of the plurality of first users to form at least one discrete feature, wherein the continuous feature comprises a continuous And generating a plurality of test features according to the at least one discrete feature and the plurality of first features, wherein the plurality of first features include an article category, a login source URL, an article browsing information, and a previous login to the blog website Website information; calculating a plurality of male ratios corresponding to a plurality of test characteristics and a plurality of female proportions; and substituting a plurality of second features when a second user browses the blog website to the classifier Calculating one of the second users based on the plurality of male ratios and the proportion of the plurality of females In addition, the plurality of second features include the continuous feature corresponding to the second user; wherein the plurality of first users and the plurality of second users continuously browse the webpage through the blog website twice Above, and the classifier is a Naïve Bayes Classifier. For example, the gender prediction method of claim 1 includes the use of a Laplace Smoothing to normalize the male ratio. Example and the proportion of the woman. For example, the gender prediction method of claim 1 of the patent scope, wherein the login source URL includes a country information or an organization information. For example, the gender prediction method of claim 1 of the patent scope, wherein the article browsing information includes an author information. For example, the gender prediction method of claim 1 of the patent scope, wherein the continuous feature is one time to browse the blog website. For example, the gender prediction method of claim 1 further includes using a hypothesis test method to reduce the number of the plurality of first features, wherein the hypothesis test method is a chi-square test. For example, in the gender prediction method of claim 1, wherein the previous website information includes an external web page different from one of the blog websites and another web page within the blog website. For example, the gender prediction method of claim 1 includes an article that dynamically changes the content, layout, and recommendation of the advertisement in the blog website according to the gender ratio.