基于扩展BP网络的城市道路造价研究
发布时间：2018-10-05 10:17
【摘要】：城市道路建设是市政工程的重要组成部分,对日常生活和经济发展具有深远影响。近年来,国家对城市道路建设的重视程度不断提高,对城市道路建设的投资也越来越多。然而,随着大批城市道路工程雨后春笋般的涌现,相应的工程管理却没有跟上时代的步伐,造价失控的现象愈发严重。特别是那些大型的道路工程,本身结构复杂,不易估算工期,加之受多种因素影响,造价计算十分繁琐,具有很大的模糊性和不确定性,难免在投资过程中出现资金浪费或者资金不足的现象,给国家和单位造成了巨大的损失,给经济发展也带来了不利的影响。要控制道路工程的投资风险,就必须准确的估算造价。针对这一问题,很多国内外学者提出了相应的解决方案,建立了多种数学模型,如BCIS法、蒙特卡罗随机模拟估算模型等。然而,这些方法大都凭借工程经验来估算道路工程成本,具有一定的主观性,难以准确计算道路工程造价。为解决上述问题,本文提出基于扩展BP网络的城市道路造价预测方法。神经网络在道路工程造价的预测中有三个问题难以解决,一是影响造价的因素难以确定,二是传统BP网络的精度和稳定性有待提高,三是训练样本和检验样本的质量难以保证。针对第一个问题,本文利用每种影响因素的贡献度来决定该因素的保留与否,即利用BP网络计算每种因素对造价误差的影响,然后确定该因素对造价计算是否具有贡献。针对第二个问题,本文拟采用扩展BP网络,用新的网络模型和混合训练算法代替传统的算法,提高道路工程造价的准确性和稳定性。针对第三个问题,本文采用欧氏距离法和k-means算法消除冗余数据和误差较大的数据。本文最后设计了一组实验,收集了2012—2014年济南市部分城区市政道路工程的部分信息,将这些信息分成训练样本和检验样本对本文算法进行验证。最终的实验结果表明,本文的方法是切实可行的,进一步提高了道路工程造价预测的准确性,对成本控制、降低工程风险具有重要的现实意义。
[Abstract]:Urban road construction is an important part of municipal engineering, which has a profound impact on daily life and economic development. In recent years, more and more attention has been paid to urban road construction and more investment has been made in urban road construction. However, with a large number of urban road projects springing up, the corresponding engineering management has not kept up with the pace of the times, and the phenomenon of out-of-control cost is becoming more and more serious. Especially those large-scale road projects, which are complicated in structure and difficult to estimate the duration of the project, together with the influence of many factors, the cost calculation is very complicated, with great fuzziness and uncertainty. It is inevitable that the phenomenon of capital waste or lack of funds in the process of investment has caused great losses to the country and the unit, and has also brought adverse effects to the economic development. In order to control the investment risk of road engineering, it is necessary to estimate the cost accurately. In order to solve this problem, many scholars at home and abroad have put forward corresponding solutions and established various mathematical models, such as BCIS method, Monte Carlo stochastic simulation model and so on. However, most of these methods rely on engineering experience to estimate the cost of road engineering, which is subjective and difficult to accurately calculate the cost of road engineering. In order to solve the above problems, this paper presents a method of urban road cost prediction based on extended BP network. It is difficult to solve three problems in the prediction of road engineering cost by neural network. One is that the factors influencing the cost are difficult to determine; the other is the accuracy and stability of the traditional BP network need to be improved; the third is the quality of the training samples and the test samples is difficult to guarantee. In order to solve the first problem, the contribution degree of each factor is used to determine whether the factor is retained or not, that is, the influence of each factor on the cost error is calculated by using BP network, and then the contribution of the factor to the cost calculation is determined. In order to improve the accuracy and stability of road engineering cost, this paper proposes to use extended BP network to replace the traditional algorithm with new network model and hybrid training algorithm. To solve the third problem, Euclidean distance method and k-means algorithm are used to eliminate redundant data and large error data. Finally, a set of experiments are designed to collect some information of municipal road engineering in Jinan from 2012 to 2014. The information is divided into training samples and test samples to verify the algorithm. The final experimental results show that the proposed method is feasible, and further improves the accuracy of road engineering cost prediction, and has important practical significance for cost control and reduction of engineering risk.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U415.13
本文编号：2253009
[Abstract]:Urban road construction is an important part of municipal engineering, which has a profound impact on daily life and economic development. In recent years, more and more attention has been paid to urban road construction and more investment has been made in urban road construction. However, with a large number of urban road projects springing up, the corresponding engineering management has not kept up with the pace of the times, and the phenomenon of out-of-control cost is becoming more and more serious. Especially those large-scale road projects, which are complicated in structure and difficult to estimate the duration of the project, together with the influence of many factors, the cost calculation is very complicated, with great fuzziness and uncertainty. It is inevitable that the phenomenon of capital waste or lack of funds in the process of investment has caused great losses to the country and the unit, and has also brought adverse effects to the economic development. In order to control the investment risk of road engineering, it is necessary to estimate the cost accurately. In order to solve this problem, many scholars at home and abroad have put forward corresponding solutions and established various mathematical models, such as BCIS method, Monte Carlo stochastic simulation model and so on. However, most of these methods rely on engineering experience to estimate the cost of road engineering, which is subjective and difficult to accurately calculate the cost of road engineering. In order to solve the above problems, this paper presents a method of urban road cost prediction based on extended BP network. It is difficult to solve three problems in the prediction of road engineering cost by neural network. One is that the factors influencing the cost are difficult to determine; the other is the accuracy and stability of the traditional BP network need to be improved; the third is the quality of the training samples and the test samples is difficult to guarantee. In order to solve the first problem, the contribution degree of each factor is used to determine whether the factor is retained or not, that is, the influence of each factor on the cost error is calculated by using BP network, and then the contribution of the factor to the cost calculation is determined. In order to improve the accuracy and stability of road engineering cost, this paper proposes to use extended BP network to replace the traditional algorithm with new network model and hybrid training algorithm. To solve the third problem, Euclidean distance method and k-means algorithm are used to eliminate redundant data and large error data. Finally, a set of experiments are designed to collect some information of municipal road engineering in Jinan from 2012 to 2014. The information is divided into training samples and test samples to verify the algorithm. The final experimental results show that the proposed method is feasible, and further improves the accuracy of road engineering cost prediction, and has important practical significance for cost control and reduction of engineering risk.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U415.13
【参考文献】
相关期刊论文 前1条
1 秦建华,李智;智能蚁群算法在化工过程优化中的应用[J];化工自动化及仪表;2005年03期
相关硕士学位论文 前1条
1 唐磊;BP网络结构确定算法的研究及仿真[D];中国石油大学;2008年
，本文编号：2253009
本文链接：https://www.wllwen.com/kejilunwen/daoluqiaoliang/2253009.html
