Migration strategies of agent based on correlation analysis and neural network

A migration model based on accumulated knowledge and the key algorithms were proposed. By correlation analysis between the executed time and the host parameters, we chose the host parameters whose correlation coefficients were great as the input of neural network and the executed time as output to construct the model of neural network. With the model, agents can evaluate the performances-of the host computers according to their parameters. Thus, the agent can choose the computer with the best performance as a running platform. The simulation experiments with aglet show that the algorithms are efficient.