A Study on Multidimensional Medical Data Processing Based on Random Forest

Medical detection is one of the important methods to prevent, diagnose and treat diseases. Under normal conditions, there are many indicators as the basis for diagnosis in medical detection usually. However, in some situations, many detection indicators, useful or important, account for a small proportion, which causes a certain cost. On the other hand, so many indicators also give inexperienced researchers difficulty in making precise decisions on the diagnosis of disease status based on more important indicators. We propose a method of multidimensional data processing based on random forest in this paper, aiming to reduce the difficulties in medical multidimensional data. We proposed a method based on Random Forest according to impact score, to classify multi-dimensional attributes as strong impact and weak impact for disease. The experimental dataset is diabetic retinopathy from the UIC. In the experiment, we designed a method based on random forest according to impact score, to classify multi-dimensional attributes as strong impact and weak impact for disease. The experimental result shows that the higher-score group has better performance in diagnosing diabetic retinopathy.