An intelligent forecasting model based on robust wavelet ν-support vector machine

摘要: Aiming at the problem of small samples, season character, nonlinearity, randomicity and fuzziness in product demand series, existing support vector kernel does not approach random curve demands time series L2(Rn) space (quadratic continuous integral space). The robust loss function is also proposed to solve shortcoming e-insensitive during handling hybrid noises. A novel wavelet machine (RW ?-SVM) based on theory modified machine. Particle swarm optimization algorithm designed select optimal parameters RW ?-SVM model scope constraint permission. results application car forecasts show that forecasting effective feasible, comparison between method this paper other ones given which proves better than traditional methods.