基于流形学习的异常检测算法研究

Manifold learning-based anomaly detection algorithm

化探异常识别是成矿预测的重要依据。化探异常识别本质上是一不均衡数据的分类问题。异常识别过程中面临的主要问题是高维数据的处理问题,流形学习通过非线性降维方法实现维数约简。提出了一种基于流形学习的异常识别算法,通过流形学习进行维数约简,结合AdaCost技术,以改善不平衡数据的分类性能。以某锡铜多金属矿床的数据为研究对象进行仿真实验,实验结果表明该算法能够更准确地圈定区域化探异常,为成矿预测与评价提供了新的解决途径。

Anomaly detection has important significance in many fields. Essentially speaking, the recognition of geochemical anomalies is the problem of imbalanced data classification. The main problems faced by anomaly identification is the processing problems of high-dimensional data, manifold learning is a nonlinear dimensionality reduction method that can reasonably reduce the data dimension. Therefore this paper proposes an anomaly detection algorithm based on the manifold learning, through mani- fold learning to achieve the dimension reduction, the new algorithm combines AdaCost technology of integrated learning, to im- prove classification performance. The new algorithm is based on the simulation experiment on the research objection of polyme- tallic deposits such as tin and copper from Gejiu, Yunnan province. The experimental results show that predicted results for the new algorithm delineating regional geochemical anomalies are better than traditional methods, which can more accurately identify the forming-ore abnormality.