The purpose of this study was to classify scrolling strategies in digital reading and to develop an algorithm to discriminate the scrolling strategies. In order to develop an algorithm for discriminating scrolling strategies, we conducted a reading experiment to record screen location and number of finger touches in 20 participants with normal vision. The reading stimuli were set at two levels of magnification that required horizontal scrolling. Two types of scrolling strategies were observed in both low and high magnification conditions : "Point Swipe," where participants had shorter finger contact times, and "Wide Swipe," where participants had longer finger contact times. As for the discriminative model of the scrolling strategy, the discriminative algorithm using support vector machines showed a high success rate in both the low and high magnification conditions. By using the findings obtained from these results, we can estimate effects of scrolling strategies on reading efficiency in digital reading for people who require magnification.