Peers' attribute data prediction in a p2p network based on the Grey Prediction Model and its improvements

A Prediction Model and two improved models are proposed and analyzed in this paper, aiming to solving the task of peers' attribute data prediction in a p2p network. The simulation results demonstrate that the Basic Grey Prediction Model performs well under the condition that the accumulation of Raw Data Series complies with the positive or negative growth of an exponential function. When this condition cannot be satisfied, the Residual Error Improvement Model can significantly increase the accuracy of the prediction. Additionally, the Grey Prediction Model with New Dynamic Information of Equal Dimension is suitable when the long term prediction is concerned.