We proposed two methods to measure the similarity of normal cloud models.One uses the expectation curves to reflect the overall feature of cloud models and to calculate the similarity by the expectation curves. The other uses the maximum boundary curve to compute the similarity between different clouds.The two methods can obtain a qualitative result

which overcomes the traditional deficiencies of the high time complexity

unstable result and excessively remarkable expectation character.The experimental results demonstrate that our methods can calculate the similarity of cloud models objectively and improve the efficiency of the algorithms in collaborative filtering recommendation and time series classification.