Logo: to the web site of Uppsala University

In recent years, Digital Twins have penetrated into the medical field, bringing revolutionary changes to the medical field. In this study, we propose a disease diagnosis algorithm, Factorization Machine Combine Product-based Neural Network (FMCPNN), which is improved on the basis of Product-based Neural Network (PNN). PNN is an end-to-end Factorization Machine algorithm, which can solve the problem of data sparseness. But PNN lacks low-order feature interaction, resulting in weak generalization ability. FMCPNN adds the second-order interaction part of FM on the basis of PNN, which improves the performance of PNN. FMCPNN can be well applied in the Digital Twins medical system to improve the accuracy and speed of disease diagnosis. Our tests show that the performance of FMCPNN surpasses some advanced models.