The neural network is a kind of the commonly used method of data mining, principal component analysis method is a kind of method that analyzes internal relationship between the many variables of the multivariate analysis of statistical. Combined the principal component analysis pretreatment method with neural network, you can analyze the relationship between the original variables, reduce dimensions of the original data and reduce the scale of data. This paper does research on the neural network algorithm and the principal component analysis correlative theory. Based on this, combined with a large number of meteorological data and disease data of Beijing, we proposed an improved method of the data mining which based on principal component analysis and neural network algorithm preprocessing. Through the contrast experiment test, the combinations of the algorithm have a large degree increase in the convergence rate and forecast accuracy property.