CN-115203565-B - Cold start method and device of recommendation system, electronic equipment and storage medium

Abstract

The embodiment of the application provides a cold start method and device of a recommendation system, electronic equipment and a storage medium. The method comprises the steps of determining a historical click content list corresponding to a user tag according to acquired user buried point data and the user tag, calculating to obtain characteristic data of the historical click content according to the user buried point data, performing smooth calculation on the characteristic data to obtain click rate data corresponding to the historical click content, determining a first candidate content list corresponding to the user tag according to the click rate data, determining a second candidate content list according to the first candidate content lists corresponding to a plurality of user tags under the same user, and performing Thompson sampling on the click rate data of the candidate content in the second candidate content list to obtain random numbers corresponding to the candidate content. And reordering the candidate contents in the second candidate content list according to the random number, and determining a user recommendation list. The embodiment of the application can provide recommended content with higher distinction for the user.

Inventors

• YI MING

Assignees

• 中国平安人寿保险股份有限公司
• 中国平安人寿保险股份有限公司

Dates

Publication Date

20260421

Application Date

20220721

Priority Date

20220721

Claims (9)

1. 1. A method for cold start of a recommendation system, the method comprising: Determining a historical click content list corresponding to the user tag according to the acquired user buried point data and the user tag; wherein, the history click content list comprises a plurality of history click contents; calculating to obtain characteristic data of the historical click content according to the user buried point data; Carrying out smooth calculation on the characteristic data to obtain click rate data corresponding to the historical click content; Determining a first candidate content list corresponding to the user tag according to the click rate data; Wherein the first candidate content list comprises a plurality of historical click contents which are arranged according to the click rate data sequence; determining a second candidate content list according to the first candidate content list corresponding to a plurality of user labels of the same user; wherein the second candidate content list comprises a plurality of candidate contents; performing Thompson sampling on the click rate data of the candidate content in the second candidate content list, and calculating to obtain a random number corresponding to the candidate content; reordering the candidate contents in the second candidate content list according to the sequence of the random numbers, and determining a user recommendation list; wherein the user recommendation list comprises a plurality of recommendation contents; the determining a second candidate content list according to the first candidate content list corresponding to the user labels of the same user includes: according to the click rate data corresponding to the historical click content in the first candidate content list, the historical click content is arranged in a descending order to obtain a list to be selected; determining the first N historical click contents in the to-be-selected list as the candidate contents to obtain a second candidate content list; wherein N is a positive integer less than or equal to a preset list threshold.
2. 2. The method of claim 1, wherein the characteristic data includes click volume data and exposure volume data.
3. 3. The cold start method of the recommendation system according to claim 2, wherein the performing a smoothing calculation on the feature data to obtain click rate data corresponding to the historical click content includes: According to the click rate data, calculating to obtain the average click rate of each historical click content in the historical click content list; According to the exposure data, calculating and obtaining the average exposure of each historical click content in the historical click content list; And carrying out smooth calculation according to the click rate data, the exposure data, the average click rate and the average exposure, and calculating to obtain the click rate data.
4. 4. The cold start method of a recommendation system according to claim 1, characterized in that the method further comprises: And if the different first candidate content lists contain the same historical click content, determining the sequence of the historical click content in the to-be-selected list according to the highest click rate data corresponding to the historical click content.
5. 5. The method for cold start of a recommendation system according to claim 1, wherein the step of performing thompson sampling on the click rate data of the candidate content in the second candidate content list to calculate a random number corresponding to the candidate content includes: calculating to obtain a click rate average value according to the click rate data corresponding to all the historical click contents in the second candidate content list; Determining a first smoothing parameter and a second smoothing parameter according to the click rate average value; Determining a first parameter and a second parameter in the thompson sampling according to the first smoothing parameter and the second smoothing parameter; And calculating the random number corresponding to the candidate content according to the first parameter and the second parameter.
6. 6. The method according to any one of claims 1-5, wherein prior to the step of thompson sampling the click rate data of the candidate content in the second candidate content list, the method further comprises: and normalizing the click rate data of the candidate content in the second candidate content list.
7. 7. A cold start apparatus for a recommendation system, the apparatus comprising: The first module is used for determining a historical click content list corresponding to the user tag according to the acquired user buried point data and the user tag; wherein, the history click content list comprises a plurality of history click contents; The second module is used for calculating and obtaining the characteristic data of the historical click content according to the user buried point data; The third module is used for carrying out smooth calculation on the characteristic data to obtain click rate data corresponding to the historical click content through calculation; A fourth module, configured to determine a first candidate content list corresponding to the user tag according to the click rate data; Wherein the first candidate content list comprises a plurality of historical click contents which are arranged according to the click rate data sequence; A fifth module, configured to determine a second candidate content list according to the first candidate content lists corresponding to the plurality of user tags under the same user; wherein the second candidate content list comprises a plurality of candidate contents; A sixth module, configured to perform thompson sampling on the click rate data of the candidate content in the second candidate content list, and calculate a random number corresponding to the candidate content; A seventh module, configured to reorder the candidate contents in the second candidate content list according to the order of the random numbers, and determine a user recommendation list; wherein the user recommendation list comprises a plurality of recommendation contents; The fifth module is specifically configured to: according to the click rate data corresponding to the historical click content in the first candidate content list, the historical click content is arranged in a descending order to obtain a list to be selected; determining the first N historical click contents in the to-be-selected list as the candidate contents to obtain a second candidate content list; wherein N is a positive integer less than or equal to a preset list threshold.
8. 8. An electronic device comprising a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for enabling a connected communication between the processor and the memory, the program when executed by the processor implementing the steps of the method according to any of claims 1 to 6.
9. 9. A storage medium, which is a computer-readable storage medium, for computer-readable storage, characterized in that the storage medium stores one or more programs executable by one or more processors to implement the steps of the method of any one of claims 1 to 6.

Description

Cold start method and device of recommendation system, electronic equipment and storage medium Technical Field The present application relates to the field of computer technologies, and in particular, to a cold start method and apparatus for a recommendation system, an electronic device, and a storage medium. Background At present, in a cold start process of a recommendation system, a recommendation method based on a user tag may cause high repeatability of recommendation lists displayed to different users, and influence user experience. Therefore, how to improve the richness of the recommended content for the user in the cold start stage of the recommendation system becomes a technical problem to be solved urgently. Disclosure of Invention The embodiment of the application mainly aims to provide a cold start method and device of a recommendation system, electronic equipment and a storage medium, and aims to improve the richness of recommended contents for users in the cold start process of the recommendation system. In order to achieve the above object, a cold start method of a recommendation system is provided according to a first aspect of the embodiments of the present application, wherein the method includes determining a historical click content list corresponding to a user tag according to acquired user buried point data and the user tag, wherein the historical click content list includes a plurality of historical click contents, calculating feature data of the historical click contents according to the user buried point data, performing smooth calculation on the feature data to obtain click rate data corresponding to the historical click contents, determining a first candidate content list corresponding to the user tag according to the click rate data, wherein the first candidate content list includes a plurality of historical click contents arranged according to the click rate data sequence, determining a second candidate content list according to the first candidate content list corresponding to the user tag under the same user, wherein the second candidate content list includes a plurality of candidate contents, performing a census calculation on the click rate data of the second candidate content list to obtain click rate data corresponding to the historical click content, determining a random number of the candidate soup content according to the random number of the candidate soup content, and determining a recommendation list according to the recommendation sequence. In some embodiments, the feature data includes click volume data and exposure volume data. In some embodiments, the calculating the characteristic data to obtain click rate data corresponding to the historical click content includes calculating an average click rate of each historical click content in the historical click content list according to the click rate data, calculating an average exposure of each historical click content in the historical click content list according to the exposure data, and calculating the click rate data according to the click rate data, the exposure data, the average click rate and the average exposure. In some embodiments, the determining a second candidate content list according to the first candidate content lists corresponding to the user labels of the same user includes arranging the historical click contents in descending order according to the click rate data corresponding to the historical click contents in the first candidate content list to obtain a to-be-selected list, determining the first N historical click contents in the to-be-selected list as the candidate contents to obtain the second candidate content list, wherein N is a positive integer less than or equal to a preset list threshold value. In some embodiments, the method further includes determining the order of the historical click contents in the to-be-selected list according to the highest click rate data corresponding to the historical click contents if the same historical click content is contained in different first candidate content lists. In some embodiments, the step of performing thompson sampling on the click rate data of the candidate content in the second candidate content list to obtain a random number corresponding to the candidate content includes calculating a click rate average value according to the click rate data corresponding to all the historical click content in the second candidate content list, determining a first smoothing parameter and a second smoothing parameter according to the click rate average value, determining a first parameter and a second parameter in thompson sampling according to the first smoothing parameter and the second smoothing parameter, and calculating the random number corresponding to the candidate content according to the first parameter and the second parameter. In some embodiments, prior to the step of Thompson sampling the click rate data of the candidate content in the second