摘要
研究了基于结构风险最小化原理的支持向量机方法对模式类的识别能力 ,构造了可用于多个模式类识别的级连式SVM模型。该模型易于实现 ,且能够找到模式间的最优分类超平面 ,泛化能力较高。支持向量机用于模式识别不存在局部极小值问题 ,且不需网络迭代训练 ,求解速度明显高于神经网络。该模型采用两种核函数 ,将SVM用于油藏测井解释中水淹层的识别以提取测井曲线与水淹级别之间的映射关系 ,从而实现模糊性油藏测井解释中水淹层的识别。实验结果表明 ,此方法对解决水淹层识别问题具有良好的适应性和实用性。
After researching the recognition capability of SVM(Support Vector Machine) based on structure minimization principle, we constructed a multi-layer SVM model for multi-pattern recognition, which can be easy realized. The SVM method can find out the super-plane between patterns without local minima, and need not to be trained. It is more efficient than neural network method and has high universal ability. We used this SVM method with two kernel functions to find the relation between well logging and water-flooded level in oil well logging. The results show that the model is suitable to water-flooded layer recognition.
出处
《计算机应用》
CSCD
北大核心
2004年第9期147-149,共3页
journal of Computer Applications
关键词
统计学习理论
支持向量机
机器学习
模式识别
水淹层识别
statistical learning theory
Support Vector Machine(SVM)
machine learning
pattern recognition
water flooded layer recognition