A Recommendation Method for Family Group Users
A recommendation method and technology for group users, applied in the field of recommendation for home group users, can solve problems such as weakening the time gap
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0053] A recommendation method for family group users, including step S4: on the basis of the dynamic time series recommendation algorithm TimeSVD++, establish a cycle model of period changes in a day, user u rates items within the time period t, and set the decay factor as:
[0054]
[0055] Among them, F t is the feature vector of user u in period t, F u The eigenvector of the time period with the most scoring data for user u; the relationship between the most active members of the family group users and their corresponding movie viewing time period is included in the user bias and user recessive factor settings, and the scoring data that is not in the active time period is calculated according to Time period similarity is biased.
[0056] The model of the present invention will score the parameter (b u , b i ,p u ) is transformed into a periodic function with days as the periodic change. At the same time, the present invention takes into account the most active memb...
Embodiment 2
[0058] This embodiment is optimized on the basis of Embodiment 1, calculates the deviation value according to the feature vector of each time period and establishes a periodic function to simulate the user interest change process, and predicts the user's rating of the item at different time periods; mainly includes the following steps:
[0059] S4-1: Establish a bias matrix factorization model; p u The last part is the second set of item factors added considering the impact of user historical behavior on rating prediction, the rating of user u on item i:
[0060]
[0061] S4-2: The popularity of the item will change with the time period due to the different viewing members at each time period of the day, and the item will be biased by b i Set to a periodic function b that varies over time periods i (t)=b i +b i,t , b i represents the fixed base value part of the item bias, and b i,t Represents the part where the item's bias changes with the time period, and the item i ...
Embodiment 3
[0075] This embodiment is optimized on the basis of embodiment 1 or 2, as figure 1 As shown, the following steps are also included before step S4:
[0076] Step S1: Estimate the user's implicit rating of the item according to the user's viewing record;
[0077] Step S2: According to the implicit rating, the rating tensor of each user for each item in each time period of the day is established;
[0078] Step S3: Use high-order singular value decomposition to decompose the rating tensor of a single user, and obtain the feature vector of the user in each time period.
[0079] The model of the present invention will score the parameter (b u , b i ,p u ) is transformed into a periodic function with days as the periodic change. At the same time, the present invention takes into account the most active members of the family group and their corresponding viewing time periods, and the score data not in active time periods should be subjected to attenuation processing. The model o...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More