计算机工程与应用 ›› 2018, Vol. 54 ›› Issue (24): 260-265.DOI: 10.3778/j.issn.1002-8331.1709-0084

• 工程与应用 • 上一篇    下一篇

基于模糊熵的KNN分类模型在岩性识别中的应用

赵彤彤1,张春雷2,张春雨3,高世臣1   

  1. 1.中国地质大学(北京) 数理学院,北京 100083 
    2.北京中地润德石油科技有限公司,北京 100083
    3.中国石油 长庆油田分公司 第四采气厂,西安 710000
  • 出版日期:2018-12-15 发布日期:2018-12-14

Application of KNN classification model based on fuzzy entropy in lithology recognition

ZHAO Tongtong1, ZHANG Chunlei2, ZHANG Chunyu3, GAO Shichen1   

  1. 1.School of Science, China University of Geosciences, Beijing 100083, China
    2.Beijing Zhongdi Runde Petroleum Technology Co., Ltd., Beijing 100083, China
    3.No.4 Gas Production Plant, Changqing Oilfield Company, PetroChina, Xi’an 710000, China
  • Online:2018-12-15 Published:2018-12-14

摘要: KNN分类模型是一种简单直接的惰性分类算法,适用于多分类问题,可应用于复杂岩性识别中。该研究以苏里格气田苏东某区为研究工区,该地区岩性结构复杂多样,其识别是本次研究工作的重点。传统KNN方法在类重叠度高的部分易判错,样本容量小的类域易误分,稀疏类的边缘点易受干扰,分类效果欠佳。为克服缺点,提出了基于模糊熵的KNN分类模型,又称为FE-KNN(Fuzzy Entropy-KNN)。FE-KNN分类模型将传统KNN与模糊理论相结合,区别对待不同特征和不同样本点,使分类的精度由84.7%提高至86.9%,为复杂碳酸盐岩岩性识别提供了一种新的思路。

关键词: 模糊熵, 特征权重, 隶属度, 样本可信度, 碳酸盐岩岩性识别

Abstract: The KNN classification model is a simple and direct inert classification algorithm, which is suitable for multi-classification problem and can be applied to complex lithology identification. The southeastern area of Sulige gas field is the research area in this study, and the lithology structure of this area is complex and diverse. Its identification is the focus of this research work. The traditional KNN method is easy to misjudge in the high degree of overlap, and the class domain with small sample size is easy to be misinterpreted. The edge of the sparse class is susceptible to interference, so that the result of classification is ineffective. In order to overcome the shortcomings, the KNN classification model based on fuzzy entropy is proposed, which is called Fuzzy Entropy-KNN(FE-KNN). The FE-KNN classification model combines the traditional KNN with the fuzzy theory, discrimination different characteristics and different sample points. The accuracy of the classification is improved from 84.7% to 86.9%. FE-KNN provides a new idea of complex carbonate rock lithology identification.

Key words: fuzzy entropy, attribute weights, membership degree, sample confidence, carbonate lithology identification