论文部分内容阅读
标准支持向量机(SVM)抗噪声能力不强,当训练样本中存在有噪声或者野点时,会影响最优分类面的产生,最终导致分类结果出现偏差。针对这一问题,提出了一种考虑最小包围球的加权支持向量机(WSVM),给每个样本点赋予不同的权值,以此来降低噪声或野点对分类结果的影响。对江汉油田某区块的oilsk81,oilsk83和oilsk85三口油井的测井数据进行交叉验证,其中核函数采用了线性、指数和RBF这3种不同的核函数。测试结果显示,无论是在SVM还是在WSVM中,核函数选择RBF识别率都是最高的,同时提出的WSVM不受核函数的影响,识别稳定性好,且在交叉验证中识别率都能够达到100%。
Standard Support Vector Machine (SVM) anti-noise ability is not strong, when the training samples in the presence of noise or outliers, it will affect the production of the optimal classification surface, eventually leading to classification results deviation. To solve this problem, we propose a weighted support vector machine (WSVM) that considers the minimum bounding sphere and assign different weight values to each sample point, so as to reduce the influence of noises or outliers on the classification results. The logging data of three wells, oilk81, oilsk83 and oilsk85, of a block of Jianghan Oilfield are cross-validated. The kernel function uses three different kernel functions: linear, index and RBF. The test results show that both the SVR and the WSVM have the highest RBF recognition rate, and the proposed WSVM is not affected by the kernel function. The recognition stability is good and the recognition rate in cross-validation can be achieved 100%.