Research On And Application Of An Intelligent Algorithm For Marriage Candidate Recommendation

With the explosive growth of user data on the Internet and the rapid expansion of machine learning in various fields,related intelligent recommendation algorithms for solving marriage problems are becoming increasingly active.This thesis takes the research of intelligent recommendation algorithms in marriage systems as the subject,focusing on the research on marriage recommendation algorithms with high accuracy,generalization ability,and robustness,and solutions to the user's cold start problem in marriage systems and combine algorithm research and application in the form of a website to recommend a group of candidates for users.The main work of this thesis is divided into three parts:1.This thesis proposes an SPWB algorithm with higher accuracy,generalization ability,and robustness.SPWB is a hybrid marriage recommendation algorithm with three algorithms stacking up,where P represents a marriage recommendation algorithm based on Popularity,W represents a marriage recommendation algorithm based on Wilson interval,and B represents a marriage recommendation algorithm based on Bayesian.This thesis uses NDCG to evaluate the recommendation algorithm,according to the experimental results,it is found that the recommendation ability of the SPWB algorithm is 5.9% higher than the NDCG value of a single marriage recommendation algorithm,and the NDCG value of the marriage recommendation algorithm based on Popularity,Wilson interval,and Bayesian is increased by 6.3%,5.7%,and 5.8% respectively.2.A solution to the user's cold start problem in the marriage recommendation system is proposed.Because the SPWB algorithm in this thesis analyzes the user's mate preferences based on the user's interactive behavior in the system and then recommends candidates for the user.But new users have no interactive behavior,and the system still needs to recommend personalized candidates for new users.Therefore,this thesis proposes the KD-KNN-LR(Combination of KD-KNN and Logistic Regression)algorithm to recommend candidates for newly registered users and uses user registration information to perform two-way matching to solve the user's cold start problem in the system.Experimental results show that the accuracy of the KD-KNN-LR algorithm is 86%based on the test user's mate selection criteria.3.In this thesis,KD-KNN-LR and SPWB algorithms are combined and applied in the marriage system,and the personal mate selection conditions are used to filter,and personalized candidates are recommended for users.The system can use the KD-KNNLR algorithm to recommend candidates for newly registered users.After the user interacts with the system,the system records the interaction record of the logged-in user and the candidate and extracts the important information of the candidate most popular with the logged-in user to form the log-in user's personal hot spot selection criteria.Then use the hotspot selection criteria of both parties to filter the ranking list of candidates generated by the SPWB algorithm,and finally get a personalized recommendation list that meets the conditions of both parties to choose a mate.The recommendation results of the marriage system show that the higher the candidate in the recommendation list,the more satisfying the log-in user's mate preference,and the goal of intelligently recommending candidates.