A Prediction Method of Cable Crosstalk in Electronic Systems with Ensemble Learning

Crosstalk between cables can seriously impede the performance of electronic systems. Cable layouts are becoming increasingly complex, and crosstalk is influenced by many factors, so it is difficult for existing crosstalk prediction methods to accurately establish direct relationships between cable near-end crosstalk and its numerous influencing factors. In order to identify the electromagnetic compatibility (EMC) problems of electronic systems in the early stage of cable layout design to guide the optimization of cable layout, the stacking-LSRM multi-stage ensemble learning method was proposed. Using this method, a cable crosstalk prediction model with high prediction accuracy and good robustness can be quickly established programmatically. Two examples were used to verify the effectiveness of the proposed method. In each example, the overall performance of the obtained optimal ensemble learning prediction model was quite good, and the effectiveness of the proposed method was verified by comparison with the prediction models established by existing methods.