The rise of algorithmic decision-making technology has not only promoted social development, but also brought about such technical and social risks as algorithmic black box algorithmic collusion, algorithmic manipulation, etc. Therefore, algorithms must be regulated to control risk. Algorithm transparency is the premise and foundation of algorithm regulation, however, due to technical and institutional barriers, complete algorithmic transparency is unrealistic and unreasonable. In this sense, it is meaningful to pursue limited and appropriate algorithm transparency in specific scenarios. Administrative regulation has unique advantages in the implementation and guarantee of algorithm transparency. Therefore, the balance and coordination between scientific and technological innovation and social security can be achieved through four specific administrative regulatory measures. The first measure is to establish a rigid and flexible administrative regulation system. The second is to promote the compliance supervision of enterprise algorithms. The third is to promote the algorithm filing system, and the last is to investigate the responsibility of the misconducts in algorithm disclosure.