Size Estimation based on Multiple Capture-Recapture Method a comparsion study

摘要

Size estimation is widely used in computer network. In this article, considering scalability and cost, we compare the various size estimation methods which are based on multiple samples and can be used in the homogeneous environment, such as P2P environments. First, according to the different theory foundations, these methods are classified into three kinds: sample-collision method, binomial distribution method and Schnabel methods. Theory analysis and experiment results show that when the total sample size is invariable, it is better to use “big sample” for binomial distribution based method. Finally, through experiments, we find that the sample-collision method has lower cost and better estimation veracity and the Schnabel method takes the second place.