Research Article

A New Dataset Size Reduction Approach for PCA-Based Classification in OCR Application

Table 1

Recognition rate for different training dataset volumes.

Number of samples used for recognition operation by -NN Recognition time for a sample (seconds)Ratio of recognition time to initial recognition time (in percent)Accuracy

60,000 = 0.1161352 = 100%97.11%95.93%95.70%
30,000 = 0.0509572 = 43.87748%96.39%94.97%94.91%
20,000 = 0.0367815 = 31.67128%94.08%93.66%93.61%
15,000 = 0.0296006 = 25.48805%93.57%93.40%93.17%