Traditional methods in Data Preprocessing haven't made good use of the internal information among the data, so the outcome inevitably apart from the real data greatly and can't appear more information among the data. The authors make use of the concept of correlation coefficient and idea of correlation analysis. A correlation analysis method is given with concrete mathematics. The authors also remedy the lost data from data source, and prove the feasibility of the method through perform mathematical calculations on mass data. The method avails the relationships among the data and drives the data known to remedy the data unknown, and it can do with the data-preprocessing problem that only have the current data not have the historical data.