Step-by-Step Regression: A More Efficient Alternative for Polynomial Multiple Linear Regression in Stream Cube

摘要

Facing tremendous and potentially infinite stream data, it is impossible to record them entirely. Thus synopses are required to be generated timely to capture the underlying model for stream management systems. Traditionally, curve fitting through Multiple Linear Regression (MLR) is a powerful and efficient modeling tool. In order to further accelerate its processing efficiency, we propose Step-by-step Regression (SR) as a more efficient alternative. As revealed in experiments, it speeds up for more than 40 times. In addition, inspired by previous work, we integrated SR into cube environment through similar compression technique to perform online analytical processing and mining over data stream. Finally, experiments show that SR not only significantly alleviates the computation pressure on the front ends of data stream management systems, but also results in a much smaller stream cube for on line analysis and real-time surveillance.