基于数据相关性分组挖掘的建筑钢材造价成本预估模型

Cost Estimation Model of Construction Steel Based on Data Correlation Group Mining

为了解决传统模型认为建筑钢材造价成本是线性变化因素,不满足实际应用,且模型结构的确定很难,在很大程度上会导致过拟合,预估准确性较差,稳定性低。通过数据相关性分组挖掘研究建筑钢材造价成本预估模型。按照各项目特征获取历史工程项目钢材造价特征指标属性集合,通过模糊模式识别,依据就近原则对待筛选样本项目和待预估样本项目的相符程度进行判断,得到和待预估项目最相近的若干历史项目当成建立预估模型的输入样本。基于最小二乘支持向量机进行数据相关性分组挖掘,建立建筑钢材造价成本预估模型。在建立模型中,正则化参数与核函数的宽度是影响建筑钢材造价成本预估结果的主要参数,通过粒子群算法获取两个参数的最优值。把得到的参数值带入模型,重新进行训练学习,获取较优的建筑钢材造价成本预估模型。在进行实验时,选择杆塔钢材、基础钢材和接地钢材三个指标作为建立模型的输入向量,将人工神经网络模型与实践序列模型作为对比进行测试。结果表明:通过人工神经网络和时间序列模型对建筑钢材造价成本进行预估,获取的预估数据有好有坏,稳定性较低,整体预估数据误差显著高于建立模型预估误差。可见建立模型预估精度高,稳定性好。

In order to solve the problem that the traditional model considers the cost of building steel as a linear variable factor and does not satisfy the practical application,and it is difficult to determine the model structure,which will lead to over-fitting to a large extent,poor prediction accuracy and low stability.The cost prediction model of building steel is studied by data correlation grouping mining.According to the characteristics of each project,the attribute set of steel cost characteristic index of historical engineering project is obtained.Through fuzzy pattern recognition,the consistency degree of selected sample items and estimated sample items is judged according to the principle of proximity.Several historical items closest to the estimated items are obtained as input samples for establishing prediction models.Based on least squares support vector machine(LS-SVM),data correlation grouping mining is carried out,and the cost prediction model of construction steel is established.In the model,the regularization parameters and the width of the kernel function are the main parameters affecting the cost estimation results of building steel.The optimal values of the two parameters are obtained by particle swarm optimization.The obtained parameters are brought into the model,and the training and learning are carried out again to obtain a better cost prediction model for building steel.In the experiment,three indexes of tower steel,base steel and grounding steel are selected as input vectors to establish the model, and the artificial neural network model is compared with the practical sequence model totest. The results show that the cost of building steel is estimated by artificial neural network and time seriesmodel. The predicted data obtained are good or bad, and the stability is low. The error of the overall predicteddata is significantly higher than that of the established model. It can be seen that the prediction accuracy of themodel is high and the stability is good.