Predictive Queries Algorithm Based on Probability Model over Data Streams

Mining the evolving trends of an online data stream and forecasting the data values in the future can provide important support for the decision-making in many time-sensitive applications. This paper models an online data stream as a continuous state transitions process by mapping the possibly infinite stream data into finite states, and uses state transition disGraph (STG) to maintain the track of the state transactions. By studying the statistic information of the history state transitions, the future values can be predicted based on the theory of Markov chain. Extensive simulation experiments are conducted and show that the predictive performance of our method is preferable to that of the existing analogous algorithms.