用人工神经网络进行空间不完备数据的插补

修回日期: 2004-03-16

刊出日期: 2005-05-15

Complement of incomplete spatial information via RBF network

Revised Date: 16 March 2004

Publish Date: 15 May 2005

摘要: 在地学研究中,特别是区域性资料处理过程中,常常遇到“不完备数据”的问题,即所谓的“数据不全”。在尽量减小估计误差的条件下对缺失数据进行预测或插补,对于充分利用历史资料和已知信息,提高预测质量具有重要意义。利用径向基人工神经网络(RBF)同时具有自组织神经网络和回归网络的优点,可以对缺失数据进行预测。实际区域地球化学数据处理的结果表明,RBF网络对空间不完备数据的建模和预测具有优异的效果。
- 空间不完备数据 /
- 人工神经网络 /
- 非线性 /
- 知识挖掘 /
- 数学地质 /
- 区域化探
Abstract: Incomplete information" and its complement are encountered frequently in geo-information processing. It is of great significance to interpolate the lost data via the known historic datasets and improve the quality and accomplishment of information integration. The RBF network possesses the advantages of Kohonen and regression networks. A test was performed to prove the effectiveness of RBF to complement the incomplete spatial information.
- incomplete spatial information /
- rule-based frame (RBF) /
- artificial neural network /
- non-linearity /
- knowledge excavation /
- mathematical geology