The main disadvantage in most prototype reduction algorithms is the excessive computational cost
especially when the prototype size is large. To deal with the problem, we present a new prototype reduction
method in which an artificial endocrine system is embedded. The method remains only for points on boundaries
between different classes. The amount of reduced rules of the reference set can be revised by granularity of
the lattice. The proposed method can get a consistent subset in a divide-reduce-coalesce manner, making it
more efficient and effective than other algorithms. The proposed approach has been tested using 11 different
datasets. The experiments show that the algorithm can give correct results when the size of dataset is large.