基于Isomap的SMO算法及在煤与瓦斯突出预测中的应用

煤与瓦斯突出发生的内在机理复杂,突出影响因素与突出事件之间的相关规律具有不确定性、模糊性,使得基于经验的传统预测方法和基于数学建模的统计预测方法的应用受到很大限制.在研究非线性降维等距特征映射和序贯最小优化算法的基础上,提出一种基于等距特征映射的煤与瓦斯突出序贯最小优化算法,该方法改进了样本向量之间的距离度量,用测地距离代替传统的欧式距离,有助于挖掘高维数据内在的几何结构.实例验证表明,该算法能可靠预测煤与瓦斯突出的危险性分类,实验进一步将Isomap和主成分分析的降维结果相比较,结果显示Isomap优于传统的线性降维技术,这说明非线性降维技术在地学数据分析中具有一定的应用潜力.

A Novel SMO Algorithm Based on Isomap and Its Application in Predicating Outburst of Coal and Gas

For complex mechanics of coal and gas outburst, the uncertainty or fuzziness between outburst-factors and outburst-hazard, makes predicting methods both based on experiences and on math-models be limited. Through analyzing Isometric feature mapping (Isomap) and Sequential Minimal Optimization (SMO) algorithm, a novel method in combining SMO algorithm and Isomap technology is presented to predicate coal and gas outburst intensity. The algorithm improves the distance measurement between samples by replacing the classical Euclidean distance with the geodesic distance, which helps preserve the intrinsic geometry of the high-dimensional data. Applying the presented forecasting method to the yearly outburst data of coal and gas shows this method has better adaptive ability and can give better forecasting results. Comparison of a standard linear dimensionality reduction method, PCA, with Isomap shows that Isomap gives better performances. Therefore, nonlinear dimensionality reduction technology has potential in the analysis of high dimensional geosciences data.