Weather radar's life prediction using support vector regression

The life of weather radar represent the time period from now on to the radar failure (not working), an accurate the life prediction is not only an indispensable element of cycle management system, but also the key technologies of improving reliability, security, maintainability of weather radar. In this paper, we present a novel methodology for predicting weather radar's life. The major contributions are: (1)We convert the weather radar's life prediction problem into a multiple regression prediction problem, and present a multiple regression framework for life prediction;(2) we induce Support Vector Regression(SVR) to solve the multivariate regression problem of weather radar's life and propose a SVR-based weather radar's life predicting algorithm. (3) In order to remove redundant and ineffective components of characteristic parameters, we propose a feature subset selection SVR algorithm. Finally, simulation results demonstrate the effectiveness of the proposed prediction algorithm.