基于Bass模型的两种参数估算算法比较研究

本文介绍了新产品扩散Bass模型及模型参数估算方法，比较了这些参数估计方法的利弊，并就中国移动用户发展情况，分别采用非线性最小二乘法和遗传算法建立扩散模型，分析和比较了两种方法的结果，得出遗传算法比非线性最小二乘法更适合于Bass模型参数估计，特别是对构建处于成长期的产品扩散模型，遗传算法可以以较少的已知数据（至少4～5个以上的数据点），得出令人满意的结果，而采用非线性最小二乘法必须已知销售峰值的数据后，才能得到较好的拟合效果。

To Compare Two Kinds of Estimates on the Parameters of Bass Model

The paper introduces the structure of Bass model and all kinds of estimates on the parameters of this model from literatures, and then compares the advantages and disadvantages of these estimates for the parameters. Nonlinear Least Squares and Genetic Algorithms are chosen to estimate the diffusion model for the mobile subscribers of China in this paper, respectively. The results of this study show us that Genetic Algorithms is better than Nonlinear Least Squares for estimating the parameters of Bass model. Especially, when a new product is in the phase of growth , and only four or five data points are available, the diffusion model can fit well with Genetic Algorithms. However, Nonlinear Least Squares can get good fitness until the time at which the highest sales rate is obtained.