The random forest classifier applied in droplet fingerprint recognition
Qing Song, Xiaoou Liu, Lu Yang · 2015
In this paper, a new method based on the random forest is supported to select appropriate feature values for liquid drop fingerprint recognition, implementing exact and effective classification of liquid. The random forest algorithm has good performance of accuracy and stability. By choosing appropriate number of elements in feature subset and training the classifier with fingerprint data, final sequences of the contribution of all feature values can be achieved. Using high-weight feature values in liquid drop fingerprint recognition, a sound result can be achieved. Theoretical analysis and experimental tests prove that the random forest algorithm can select the outstanding feature values which could really embody the attribute of liquid samples. The averaged rate of recognition is high to 95.57%.