小样本装备软件成本估算相关向量机建模

摘要

Equipment software cost estimates face the problem of small sample. The paper proposes using machine learning theory for this field and constructes the associated vector machine based on a small sample of equipment software cost estimation model. Using COCOMO model database to train and validate the model built. To further verify the rationality of the model buildt,the support vector machines and neural network model are used to estimate the same data. And experimental results show that the associated vector machine model is more accurate.%针对装备软件成本估算时面临的历史数据少的小样本难题,提出把机器学习理论运用到装备软件成本估算研究领域,构建了基于相关向量机的小样本装备软件成本估箅模型,采用COCOMO模型数据库中的数据对所构建模型进行训练和验证.为进一步验证所构建模型的合理性,在机器学习领域分别运用支持向量机模型和神经网络模型对相同数据进行估算,实验结果证明了基于相关向量机估算模型在小样本装备软件成本估算时比其他两种模型更精确.