Hybrid Tree-Encoded Linear Model for Video Recommendation Cold Start
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video recommendation algorithms face challenges in addressing the cold-start problem, where new users or users with limited viewing records cannot receive personalized recommendations due to data sparsity, affecting user experience and stickiness in video services.
Innovation Solution
A method utilizing a hybrid tree-encoded linear model algorithm to predict interest similarities between users, enabling the recommendation of users with similar interests to new or infrequent users, thereby creating a personalized recommendation list.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional video recommendation algorithms (CB recommendation, CF recommendation, social recommendation) are used, then recommendation functionality is provided, but the cold-start problem and data sparsity problem cannot be solved, resulting in poor performance for new users or users with less behaviors
Solution Approach 1:
The patent segments users into different types based on their behavior data availability: users with sufficient viewing records and users with insufficient viewing records (cold-start users). Different recommendation strategies are applied to each segment - conventional algorithms for users with data and the patent's innovative approach for cold-start users, thereby resolving the contradiction between recommendation performance and applicability to new users
Solution Approach 2:
The patent changes the parameter of user data requirements by introducing a threshold mechanism. Users are classified based on whether their viewing record count exceeds a predefined threshold. This parameter change allows the system to adaptively select different recommendation approaches, solving the contradiction by making the system versatile for both new and experienced users while maintaining reliability through appropriate algorithm selection
2Adaptability or versatility
If user clustering methods are used to resolve cold-start problem (searching for users of similar age, same gender, geographical proximity, or same community), then cold-start problem is addressed, but the solution lacks precision in predicting actual interest similarities
Solution Approach 1:
The patent introduces an interest similarity prediction model as an intermediary between user demographic/behavioral data and recommendation generation. This model precisely calculates interest similarity scores based on viewing records, video attributes, and user profiles, thereby improving measurement precision of interest similarity while maintaining adaptability to cold-start users through alternative data sources
Solution Approach 2:
The patent replaces the mechanical clustering approach (grouping by demographics/geography) with a data-driven similarity prediction system that uses machine learning models to calculate interest similarity. This substitution improves measurement precision by using actual viewing behavior and video attribute matching rather than rough demographic proxies, while still addressing cold-start through alternative data sources
Data Source
AI summary
This application relates to an interest recommendation method, a computer device, and a storage medium. The method includes: obtaining feature information of a target user; predicting interest similarities between the target user and a user group according to the feature information by using an interest similarity prediction model, the interest similarity prediction model being implemented according to an interest similarity between each pair of users in a sample set of historical records of users based on a hybrid tree-encoded linear model algorithm, the hybrid tree-encoded linear model algorithm being implemented based on a tree model and a linear model; determining, according to the interest similarities, recommended users in the user group having interests similar to those of the target user; and obtaining an interest list of the recommended users, and creating a recommendation list for the target user according to the interest list.