4.3 样本训练和预测工作实施 网络所需的样本数主要由两方面因素决定:一是网络映射关系的复杂程度(映射关系越复杂,要求提供的样本数便越多);二是数据中的噪声(所需样本数随着噪声的加大而增多)。 样本数考虑到土方开挖及支撑施工周期,选择预测前若干日的实际监测数据(设取30组)作为“样本数据对”;其中,取前25组作为学习、训练组样本, 4.3 Sample training and forecasting The number of samples required to implement the network is mainly determined by two factors: one is the complexity of the network mapping (the more complex the mapping is, the more samples are required to be provided); the other is the noise in the data (The number of samples required increases with increasing noise). The number of samples Taking earth excavation and support construction period into consideration, the actual monitoring data (30 groups) for a few days prior to the forecast are selected as the “sample data pairs”; the first 25 groups are taken as the samples for learning and training groups,