概率神经网络的边界预报

Bound Prediction by Probabilistic Neural Network

在概率神经网络用于时间序列预测的基础上,结合数理统计理论,提出在一定置信度情况下,对时间序列的边界预报问题.在时间序列预报时,不仅给出预报值,同时也给出预测值的变化范围.通过模拟数据和齿轮箱实际数据,对边界预报进行了应用分析,得出预报具有滞后性的结论并分析了其原因;同时讨论了平滑因子等参数对边界预报的影响.

Probabilistic Neural Network (PNN) is applied successfully to the time series prediction because of its high identifiabilities. Based on the time series prediction by probabilistic neural network, the bound prediction was presented by using PNN. By analyzing the simulated data and the experimental data of a gearing box, the lag property of the prediction and the influences of some parameters on the bound prediction were obtained.