谱聚类划分随机森林算法在图像识别中的应用

随机森林是近些年发展起来的新集成学习算法,具有较好的分类准确率。针对该算法计算复杂度较高的不足,提出了一种基于谱聚类划分的随机森林算法。首先,利用聚类效果较好的谱聚类算法对原始样本集的每一类进行聚类处理。然后,在每一聚类簇中随机选取一个样本作为代表,组成新训练样本集合。最后,在新训练样本集上训练随机森林分类器。该算法通过谱聚类技术对原始样本进行了初步划分,将位置相近的多个样本用簇内的一个样本代表,较大程度地减少了训练样本的个数。在Corel Image图像识别数据集上的实验表明,算法可以用较少的分类时间达到较高的分类精度。

A normalized cut spectral clustering based random forest algorithm and its application on image recognition

Random forest is a new ensemble learning algorithm with a good classification accuracy. As its computational complexity is high, this paper proposes a normalized cut spectral clustering based random forest algorithm. First, use normalized cut spectral clustering algorithm to cluster the original sample set, then randomly select in each cluster a sample as representative and construct a new training set. Finally, train the random forest classifier on the new training sample set. Through spectral clustering techniques, the original samples can be divided. As the similar samples in the same cluster are represented by one random sample, the number of training samples are largely reduced. Experiments on the Corel Image data sets show that the proposed algorithm can achieve a higher classification accuracy with less free time.