Personalized Recommendation Method of Health Education Resources in Home-based Aged Care Based on Similarity Correlation

When new home-based health education resources are added to personalized recommendation system, it is difficult to determine preference differences due to the lack of user interaction, resulting in a cold start problem, which leads to a low degree of matching between the characteristics of recommended resources and user preferences. Therefore, a personalized recommendation method of health education resources in home care for the aged based on similarity correlation is proposed. Establish a user information table to generate portrait data of elderly health education resources preference. Based on Manhattan distance, the similarity difference of resource preference is calculated, and the preference category is refined by using the neighbor selection mechanism of similarity difference threshold, which initially improves the matching degree of personalized recommendation. Based on the results of similarity classification, a personalized association recommendation model is constructed based on the deep neural network model, and the parameters of the recommendation model are optimized by gradient descent method, so as to make personalized recommendations for health education resources in home-based care for the elderly. Considering the cold start problem, the interactive feedback optimization of personalized association recommendation is carried out based on encoder group, which further improves the matching degree between recommended resource characteristics and user needs. The test results show that the recommendation response time of this method is less than 11.31s, and the matching degree between recommended resources and user preferences is higher than 91.93, which meets the actual needs.