基于引力模型的朴素贝叶斯分类算法*

针对朴素贝叶斯分类器在分类过程中,不同类别的同一特征量之间存在相似性,易导致误分类的现象,提出基于引力模型的朴素贝叶斯分类算法。提出以引力公式中的距离变量的平方作为“相似距离”,应用引力模型来刻画特征与其所属类别之间的相似度,从而克服朴素贝叶斯分类算法容易受到条件独立假设的影响,将所有特征同质化的缺点,并能有效地避免噪声干扰,达到修正先验概率、提高分类精度的目的。对遥感图像的分类实验表明,基于引力模型的朴素贝叶斯分类算法易于实现,可操作性强,且具有更高的平均分类准确率。

Naive Bayesian Classification Algorithm Based on Gravity Model

In order to solve the problem of misclassified in the process of Naive Bayesian classifier which caused by the similarity between the same feature quantities of different categories, this paper presents a simple Bayesian classification algorithm based on gravitational model. For the purpose of overcome the influence of the Naive Bayesian classification algorithm, it is easy to be influenced by the conditional independent hypothesis, and can effectively avoid noise interference, to correct the prior probabilities, improve the accuracy of classification purposes. This paper proposes a gravitational model to describe the similarity between the feature and its category by using the square of the distance variable in the gravitational formula as the "similar distance". The classification experiments of remote sensing images show that the naive Bayesian classification algorithm based on gravitational model is easy to implement, has high operability and has higher average classification accuracy.