Research And Implementation Of Intelligent Marriage Recommendation System

With the popularization and rapid development of network information technology,the Internet has become an important way for people to date and make friends.Compared to traditional product recommendations,marriage recommendations have the characteristic of bidirectionality,where users are both the recommending subject and object.This thesis focuses on the bidirectional,accurate,and fair nature of marriage recommendations,and conducts relevant research on marriage recommendation algorithms and matching algorithms,applying them to a marriage recommendation system.The main work includes three aspects:1.A hybrid recommendation algorithm based on Stacking fusion is proposed.Firstly,basic marriage-related features and personality characteristics were extracted from the original dataset.Then,based on the processed dataset,three machine learning models for marriage prediction are built using XGBoost,Light GBM,and Cat Boost respectively.Through experiments,it was proved that introducing personality characteristics can improve the predictive performance of each model.Then,by using grid search to optimize the hyperparameters of each model,the training performance of each model was improved.Lastly,a hybrid recommendation algorithm based on Stacking fusion was proposed to further improve the predictive accuracy and generalization ability of the model.It uses Stacking to fuse the XGBoost,Light GBM,and Cat Boost models while incorporating user personality characteristics.Experiments have shown that the fused model is superior to the individual XGBoost,Light GBM,and Cat Boost models in terms of accuracy,f1＿score,and AUC.2.A non-equilibrium preference-matching algorithm based on Gale-Shapley is proposed.Firstly,the problems with applying the traditional Gale-Shapley matching algorithm to marriage recommendation systems were analyzed through experiments.Then,a non-equilibrium preference-matching algorithm based on Gale-Shapley was proposed.It combines the practical needs of marriage recommendation systems by introducing a lower bound for match scores and a balance factor for passive matching parties,as well as redefining the rules for ending matches.Through experiments,it was shown that the non-equilibrium preference-matching algorithm based on Gale-Shapley can achieve stable matching even when there is an uneven distribution between male and female users.Compared to the traditional Gale-Shapley matching algorithm,the proposed algorithm improves overall matching quality,balances the differences between active and passive matching parties,strengthens bidirectionality,and reduces time complexity.3.The hybrid recommendation algorithm and non-equilibrium preference-matching algorithm proposed are applied to a marriage recommendation system.The system uses the hybrid recommendation algorithm to predict potential marital satisfaction for users and recommend a set of candidates.Then,the system calculates the compatibility between users and candidates based on their preferences and generates a ranking of preferred candidates.Finally,the non-equilibrium preference-matching algorithm is used to perform one-to-one matchmaking among all male and female users,and recommended partners are generated based on the generated matches.