Research on null-value estimation algorithm based on predicted value

Information system composed of multi-source information is usually incomplete information system, the system contains a variety of types of null values. Incomplete information system of air-missing value estimation problem, first introduced abroad to fill the existing empty value method, and compared and analyzed; Secondly, the detailed model based on the similarity between the null value estimation methods (Expectation maximization, SIM-EM), and analyzed when dealing with sparse data problems; again, a problem of sparse data for improved forecasting methods score - the predicted value based on a null value estimation algorithm (Prediction-EM) and introduce a new feature weight method; Finally, experimental results and performance of this algorithm was validated.