Abstract: Sine Cosine Algorithm（SCA） is a novel intelligent optimization algorithm. SCA finds the best solution by the changes in the values of sine and cosine functions. The conversion parameter can balance the global exploration and local exploitation abilities of SCA. It has an important influence on the algorithm. To improve the optimization performance of SCA, the setting of conversion parameter is analyzed firstly and then SCA with parabolic function decreasing conversion parameter and SCA with exponential function decreasing conversion parameter are proposed. The numerical experiments on benchmark functions are performed to compare the presented algorithms and conversion parameter linearly decreasing SCA. The results show that SCA with exponential function decreasing conversion parameter has higher calculation accuracy and faster convergence speed. Finally, the calculation of similarity function for collaborative filtering recommendation algorithm is used to further verify the feasibility and effectiveness of the algorithm.

Key words: sine cosine algorithm, conversion parameter, global exploration and local exploitation, optimization

摘要： 正弦余弦算法是一种新型智能优化算法，利用正弦函数和余弦函数值的变化来实现优化搜索。转换参数直接影响算法全局探索和局部开发的平衡，对算法的性能有着重要影响。为提高该算法的优化性能，首先对转换参数的设置进行分析，然后设计出转换参数抛物线函数递减和指数函数递减两种正弦余弦算法，并采用标准测试函数进行数值实验，和转换参数线性递减的基本正弦余弦算法进行比较。结果表明指数函数递减的正弦余弦算法具有更高的计算精度和更快的收敛速度。最后以协同过滤推荐算法中相似度函数的计算为应用对象，进一步验证新算法的可行性和有效性。

关键词: 正弦余弦算法, 转换参数, 全局探索和局部开发, 最优化