Abstract:

To solve the problem that the choice of parameters influence the forecast accuracy of support vector regression (SVR), a SVR forecasting parameters optimization approach based on artificial bee colony algorithm is proposed. The experiment results show that the proposed approach can avoid trapping in the local minimum solution and has the higher forecasting accuracy than the particle swarm optimization algorithm. The approach is applied to the analysis of lubrication metal content time series of airborne vehicle’s powerplant under fault condition, the occurring of fault is forecasted successful.