With the acceleration of aging, more and more elderly people choose to supplement with nutritional supplements, and human pursuit of nutrition and health has never stopped. However, the selection of nutritional supplements in the current market is often too complex, and this situation can also make it difficult for many people to find products that are suitable for themselves. This article designs a personalized nutritional supplement recommendation system based on linear regression, support vector machine, and naive Bayesian methods in machine learning technology. It can automatically help users find suitable nutritional supplements in the market. By designing a system framework and applying algorithms to data collection and preprocessing, the applicability of a personalized nutrition supplement recommendation system based on machine learning technology is evaluated through data comparison. Research has found that personalized nutrition recommendation systems can effectively help consumers choose products that are suitable for them, and have a positive impact on maintaining physical health. The application of linear regression based on machine learning technology in personalized nutritional supplement recommendation systems is far more accurate than traditional methods, support vector machines, and naive Bayesian methods, with linear regression accuracy reaching 96%.