Method and device for obtaining optimal parameter combination of recommendation system
A parameter combination and recommendation system technology, applied in the field of recommendation systems, can solve problems such as large quantities, insufficient confidence in user performance data, and inability to efficiently and accurately determine the optimal parameter combination of the recommendation system
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0059] see figure 1 , figure 1 It is a flow chart of the method for obtaining the optimal parameter combination of the recommendation system in Embodiment 1 of the present invention, including the following steps:
[0060] S110: Screen a plurality of experimental users.
[0061] According to actual needs, a certain proportion of users can be randomly sampled from the total number of users as experimental users. For example, randomly sample 1% of all users as experimental users. The sampling ratio can be set according to actual requirements. The higher the sampling ratio, the greater the processing capacity of this embodiment, but the better the effect of obtaining the optimal parameter combination.
[0062] S120: Perform a parameter test at the granularity of the experimental user's session (Session) to obtain a learning sample, where the learning sample includes partial values of parameter combinations and their corresponding system effect evaluation values.
[0063] Di...
Embodiment 2
[0070] This embodiment introduces a specific implementation manner of step S120 in the first embodiment.
[0071] For ease of understanding, before introducing specific steps, two related concepts are introduced first.
[0072] First, the user's session:
[0073] In a recommendation system, user behavior patterns are usually refreshed and browsed in units of sessions. That is: starting from a certain moment, continuously refresh and browse and read articles for a period of time, then stop, and then continue to refresh and browse for a period of time after a period of time, and so on.
[0074] Let the user be u, and the i-th session of user u is recorded as s u (i). A user session can contain many refreshes, namely s u (i)={r u (i,j)|1u (i)}, where r u (i, j) is the j-th refresh in the i-th session of user u, n u (i) is the total refresh times in user u's ith session.
[0075] remember st u (i,j) is the starting time of the jth refresh in the ith session of user u, et ...
Embodiment 3
[0104] This embodiment introduces a specific implementation manner of step S140 in the first embodiment. see Figure 4 , Figure 4 It is the implementation flowchart of the third embodiment of the present invention.
[0105] In this embodiment, it is necessary to find the optimal system effect and its corresponding parameter combination value from the above system effect space, and the corresponding parameter combination value is the optimal parameter combination of the recommendation system. Since the amount of data contained in the system effect space is very large, it is impossible to find the optimal system effect through direct comparison. Therefore, this embodiment adopts the hill climbing method to obtain the optimal system effect in the system effect space. Refer to the following Figure 4 Details. Embodiment 3 of the present invention comprises the following steps:
[0106] S141: Randomly select the value P of multiple parameter combinations in the parameter comb...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More 


