基于加权组合模型的二手车价格预测方法  

Prediction Method of Used Car Price Based on Weighted Combination Model

作　　者：周远 贺波涛 ZHOU Yuan;HE Botao(Wuhan Research Institute of Posts and Telecommunications,Wuhan 430037;Wuhan FiberHome Digital Technology Co.,Ltd.,Wuhan 430037)

机构地区：[1]武汉邮电科学研究院,武汉430037 [2]武汉烽火众智数字技术责任有限公司,武汉430037

出　　处：《计算机与数字工程》2024年第5期1449-1452,1458,共5页Computer & Digital Engineering

摘　　要：为了提高二手车价格预测的精度与可靠性,提出了一种基于神经网络与贝叶斯优化LightGBM算法的加权组合方法,对具有多个影响因素的二手车价格进行预测。首先对原始数据进行数据预处理,其次对预处理后的数据集进行特征工程以获得适合神经网络与树模型训练的数据集,然后分别使用神经网络与贝叶斯优化的若干机器学习算法进行训练获得网络模型,最后将各模型进行组合并与单一模型进行对比。预测结果显示,提出的神经网络与改进LightGBM算法的加权组合模型比单一模型以及其他组合方式的模型预测能力更强。In order to improve the accuracy and reliability of used car price prediction,a weighted combination method based on neural network and Bayesian optimization LightGBM algorithm is proposed to predict the price of a used car with multiple influ-encing factors.First,data preprocessing is performed on the original data set.Secondly,feature engineering is conducted on the pre-processed data set to obtain data sets suitable for training neural networks and tree models.Then,several machine learning algo-rithms optimized by neural network and Bayesian optimization are employed for training to obtain the network model.Finally,the models are combined and compared with a single model.The prediction results show that the weighted combination model of the pro-posed neural network and the improved LightGBM algorithm has stronger predictive ability than the single model and other combina-tion models.

关 键 词：LightGBM算法 神经网络 贝叶斯优化 机器学习 组合模型预测 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]