When the causality-relationship is incomplete, it’s easy to have problem on sample classification. For the sake of solving this problem, this paper proposes an improved classification recognition algorithm based on causality analysis. This algorithm has improved the process of classification and recognition which is proposed in Causality Analysis in Factor Spaces [1], and it’s based on the nearest-neighbor rule and maximum subordination principle. In addition, aiming at the case that can be only applied in the discrete groups in Pei-Zhuang Wang’s paper, this article has transformed the continuous data into discrete data by segmentation method. Therefore, this algorithm expands on its original application into the case involving continuous data. Experimental results indicate that this improved classification recognition algorithm can successfully identify all the samples, and it also significantly improves the overall recognition rate. Simultaneously, when continuous data is centralizing, this algorithm is better than most common classification algorithms, and it can be effectively applied to image classification areas.