CN106875278A - Social network user portrait method based on random forest - Google Patents

Abstract

The present invention provides a social network user portrait method based on random forest, which specifically includes the following steps: obtaining multi-source attribute data of online social networking sites; labeling the data attribute sets of original multi-source attributes with original attribute labels, and calling similarity function pairs The data attribute sets of different attributes are traversed for similarity detection; according to the decision tree of the original single-layer multi-source attribute, the data attribute sets whose similarity meets the threshold range are merged to generate the merged attribute label, and the random forest algorithm is used to train the samples; the voting majority is obtained, Assign weights to the obtained voting majority, and then sort according to the weights from large to small to obtain all tag weight values; retain tags within the preset threshold to form a new tag attribute set for the portrait of attributes in the user's social network. The purpose of the present invention is to adopt a random forest model for user's attribute label division, which effectively improves the problem of deficiency and complexity of the traditional attribute division based on small sample sampling.

Description

Random forest based user portrait method for social network

technical field

The invention relates to the technical field of online social networks, in particular to a random forest-based social network user portrait method.

Background technique

The research on online social network is the key field of academic research in recent years. my country has the largest Internet users in the world. Therefore, a large amount of data has been generated in the early stage of Internet promotion and current use. The vast majority of data resources are idle and cannot be well processed and commercialized, causing huge losses, and it is not conducive to the further development of social networks. Major Internet companies have invested huge financial resources and manpower in the field of online social relations. It is of great significance to carry out a series of researches on the rational development and use of Internet data resources.

Contents of the invention

The present invention provides a social network user portrait method based on random forest, aiming at adopting random forest model for user's attribute label division, effectively improving the problem of deficiency and complexity of traditional attribute division based on small sample sampling.

In order to solve the above problems, an embodiment of the present invention provides a random forest-based social network user portrait method, which specifically includes the following steps:

Obtain multi-source attribute data for online social networking sites;

Label the data attribute collection of the original multi-source attribute with the original attribute label, and call the similarity function to traverse the similarity detection for the data attribute collection of different attributes;

According to the decision tree of the original single-layer multi-source attributes, the data attribute sets whose similarity meets the threshold range are merged to generate the merged attribute labels, and the random forest algorithm is used to train the samples;

Obtain the majority of votes, assign weights to the obtained majority of votes, and then sort according to the weights from large to small to obtain all tag weight values;

The tags within the preset threshold are retained to form a new tag attribute set for the portrait of attributes in the user's social network.

As an implementation mode, the following steps are also included:

Set the minimum detection termination threshold, when the similarity is less than the minimum detection termination threshold, the similarity detection of this set is terminated.

As an implementation manner, the minimum detection termination threshold is 0.15.

As an implementation manner, the similarity function is:

Among them, α is the similarity adjustment parameter, α∈[0,1], ω(x) represents two attribute functions with higher label similarity.

As an implementation manner, the value of α is 0.001.

As an implementation, the step of retaining tags within the preset threshold and forming a new tag attribute set for the portrait of attributes in the user's social network specifically includes the following steps:

Set the label mode threshold. When the vote mode obtained by the random forest algorithm is less than the label mode, the label is considered unrepresentative and the label is discarded;

The reserved tags are sorted according to the tag weight value from large to small to form a new tag attribute set.

As an implementation manner, the range of the similarity threshold is [0.9, 1].

Compared with the prior art, the present invention has the beneficial effects of adopting the random forest model for user's attribute label division, effectively improving the problem of deficiency and complexity of the traditional attribute division based on small sample sampling.

Description of drawings

FIG. 1 is a flow chart of the random forest-based social network user portrait method of the present invention.

detailed description

The above and other technical features and advantages of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Apparently, the described embodiments are only some of the embodiments of the present invention, not all of them.

As shown in the figure, a random forest-based social network user portrait method specifically includes the following steps:

S100: Obtain multi-source attribute data of online social networking sites, and import them into a data storage system;

S101: Label the data attribute set of the original multi-source attribute with the original attribute, and call the similarity function to traverse the similarity detection for the collection of different attributes. The similarity function is:

Among them, α is the similarity adjustment parameter, α∈[0,1], ω(x) represents two attribute functions with higher label similarity. However, in practice, the value of α is generally very small, and the test value dependent on the sample is constantly revised. According to the experimental results, when α is increased by an order of magnitude, the selected features are very few, and when α is reduced by an order of magnitude, the obtained value is almost unchanged. , therefore, in this embodiment, α uses 0.001;

S102: Set the lowest detection termination threshold, when the similarity is less than the lowest detection termination threshold, terminate the similarity detection of the set, wherein the minimum detection termination threshold is 0.15;

S103: According to the decision tree of the original single-layer multi-source attribute, after merging the sets whose similarity meets the threshold range to generate the merged attribute label, use the random forest algorithm to train samples, and the similarity threshold range is [0.9,1];

S104: Obtain the majority of votes, assign weights to the obtained majority of votes, and then sort according to the weights from large to small to obtain all tag weight values;

S105: Keep the tags within the preset threshold, and form a new tag attribute set for the portrait of attributes in the user's social network. The specific implementation method is: set the tag mode threshold, when the voting mode obtained by the random forest algorithm is less than the When the threshold is higher than the threshold, the label is considered unrepresentative and discarded; the reserved labels are sorted according to the label weight value from large to small to form a new label attribute set, and the new label attribute set is used for user portraits in social networks .

Compared with the prior art, the present invention has the beneficial effects of adopting the random forest model for user's attribute label division, effectively improving the problem of deficiency and complexity of the traditional attribute division based on small sample sampling.

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention and are not intended to limit the protection scope of the present invention. . In particular, for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (7)

1. A social network user portrait method based on random forest, is characterized in that, specifically comprises the steps: Obtain multi-source attribute data for online social networking sites; Label the data attribute collection of the original multi-source attribute with the original attribute label, and call the similarity function to traverse the similarity detection for the data attribute collection of different attributes; According to the decision tree of the original single-layer multi-source attributes, the data attribute sets whose similarity meets the threshold range are merged to generate the merged attribute labels, and the random forest algorithm is used to train the samples; Obtain the majority of votes, assign weights to the obtained majority of votes, and then sort according to the weights from large to small to obtain all tag weight values; The tags within the preset threshold are retained to form a new tag attribute set for the portrait of attributes in the user's social network. 2. the social network user portrait method based on random forest according to claim 1, is characterized in that, also comprises the following steps: Set the minimum detection termination threshold, when the similarity is less than the minimum detection termination threshold, the similarity detection of this set is terminated. 3. The random forest-based social network user portrait method according to claim 2, wherein the minimum detection termination threshold is 0.15. 4. the social network user portrait method based on random forest according to claim 1, is characterized in that, described similarity function is: Among them, α is the similarity adjustment parameter, α∈0,1, ω(x) represents two attribute functions with higher label similarity. 5. The random forest-based social network user portrait method according to claim 4, wherein the value of α is 0.001. 6. The social network user portrait method based on random forest according to claim 1, characterized in that, the label within the preset threshold is retained to form a new label attribute set for the portrait step of attributes in the user social network, Specifically include the following steps: Set the label mode threshold. When the vote mode obtained by the random forest algorithm is less than the label mode, the label is considered unrepresentative and the label is discarded; The reserved tags are sorted according to the tag weight value from large to small to form a new tag attribute set. 7. The random forest-based social network user portrait method according to claim 1, wherein the similarity threshold range is [0.9,1].