基于一步向前预测回归树的内燃机振动信号趋势预测

Step Forward Prediction Based on Regression Trees Vibration Signal Prediction

针对内燃机振动信号非线性、非平稳特点,提出了一种基于一步向前预测回归树的新模型。首先确定合适的嵌入维数和时间延迟,从而据此确定回归树的输入;然后,建立回归树预测模型;最后,用原始时间序列对所建立的模型进行预测性能评价,并将此方法与神经网络、支持向量机预测方法相比较。

According to the non-linear and non-stationary characteristics of combustion motor vibration signals,a new model is presented here based on the one-step-ahead prediction regression tree.First,defined the proper embedding dimension and time delay thus determined the input regression tree defined.Next,the prediction model of regression tree is built.Last,the prediction performance of the model built through the initial time series is evaluated,and compared this method with neural network method and support vector machine forecasting.