改进的决策树算法在手写体汉字识别中的应用

摘要

Basic flow of off-line handwriting Chinese character recognition was introduced Then, taking the traditional decision tree ID3 model as basis, the selection strategy of decision attributions was modi fied. Based on the computation standard of information gain and by introducing correlativity of the attribu tions, the defect of "multiple-valued bias" of traditional ID3 algorithm was supplemented by using this method. A pre-pruning strategy was proposed based on node coordination degree. The classification preci sion of the improved ID3 algorithm and the traditional ID3 algorithm was analyzed and compared with each other. Finally, the improved ID3 algorithm was used for handwriting Chinese character recognition. Ex perimental result showed that the improved ID3 algorithm could reduce training time and improve recogni tion percentage.%介绍脱机手写体汉字识别的基本流程,以传统决策树ID3模型为基础,对决策属性挑选策略进行改进.该方法基于信息增益的计算标准,通过引入属性间相关性关系对传统ID3算法“多值偏向性”的缺陷进行补偿,提出节点协调度的预剪枝策略.从分类精确度方面对改进的ID3算法和传统D3算法进行比较和分析.将改进的ID3算法应用于手写体汉字识别中,试验结果表明,改进的ID3算法可以减少训练时间,并提高识别率.