电梯价格预测中的数据预处理方法研究

Research on Methods of Data Pre-Processing for Prediction of Elevator Price

电梯参数繁多,参数中特征复杂,针对电梯价格预测时不准确的特征表示会导致预测准确率低的问题,提出一种面向电梯价格预测的数据预处理方法。首先对数据进行特征聚类,对于不同类别的特征进行不同的标准化表示;然后根据特征的信息增益选择对价格预测影响大的关键特征用于机器学习。以电梯设备为研究对象,通过将原始数据划分为不同的训练集和测试集比例,建立机器学习模型进行实验。实验结果表明,该数据预处理方法既能够加快神经网络的训练速度,还能够提高预测的准确率。

Elevator has large number of parameters with complex features in it,aiming at the problem that improper represen⁃tation of features can leads a bad accuracy in prediction of elevator price,a method of data preprocessing for prediction of elevator price is proposed.Firstly,different standardized representations are used to different categories of features after feature cluster.Sec⁃ondly,key features that have greater impact on price prediction are selected for machine learning based on information divergence of features.Taking the elevator as research object and dividing raw data into different ratios of train and test dataset is for develop⁃ing machine learning model to do experiment.The experiment result showing that this method of data preprocessing proposed can speed up the training process,and also improve the accuracy of the prediction.