Abstract: Considering the complexity of the influence factors of wax deposition and the advantage of least squares support vector machine in small sample prediction，based on the prediction principles of least squares support vector machine，by optimizing the parameters of least squares support vector machine，the prediction model of wax deposition rate was established and wax deposition rate was predicted. The method had higher accuracy when the samples were fewer，and the prediction results of wax deposition rate was in good agreement with the experimental data. The least squares support vector machine could get the intuitive function expression when it was used to establish the model of wax deposition rate，while neural network method could not get explicit expression. so this method has sufficient preponderance. When the RBF kernel function was used，different initial values of regularization parameters ? and kernel bandwidth ? had a greater impact on the predicted results，so it should be used with care.

Key words: least squares support vector machine, wax deposition rate, prediction, model, model accuracy

摘要： 考虑蜡沉积影响因素的复杂性和最小二乘支持向量机在小样本预测方面的优势，基于最小二乘支持向量机预测的原理，通过优化最小二乘支持向量机的参数，建立了蜡沉积速率的预测模型，并对蜡沉积速率进行了预测。结果表明：该方法在样本数量较小时仍具有较高的精度，蜡沉积速率的预测值和实验值的吻合程度较好；最小二乘支持向量机建模时可以得到直观的函数表达式，而神经网络方法却不能得到模型的显式表达式，因此该方法具有明显的优势；应用径向基核（RBF）作为核函数时，不同初值的正则化参数?和核函数宽度?对预测结果具有较大影响，使用时应合理选择。

关键词: 最小二乘支持向量机, 蜡沉积速率, 预测, 模型, 模型精度