基于单隐层前馈神经网络极限算法的高速公路工程投资估算模型研究

发布时间：2018-03-13 14:57

本文选题：高速公路工程　切入点：投资估算　出处：《江西理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：本文首先分析选题背景和意义,梳理国内外研究现状,针对我国目前高速公路项目工程投资估算的现状,找出当前高速公路投资估算存在的问题;详细介绍了高速公路方面的投资定义以及相关的投资估算理论,并对比了已有投资估算方法的优缺点;同时,从工程项目的投资结构、主要特征、有关文献记载等三个方面,利用灰色理论深入分析高速公路项目投资预算特征参数,建立了投资估算工程特征参数体系;将灰色关联分析法与单隐层前馈神经网络极限算法结合起来,通过对影响高速公路投资因素的全面考虑,建立相关模型,将7个工程特征参数引入高速公路概念,并将其定义为模型的自变量,基于最小二乘原理,直接将误差确定在允许范围内,各种参数不需反复地多次训练求得,最后快速估算出拟建高速公路工程的投资,利用实际数据对此算法进行科学验证,以突出其良好的应用灵活性与适应性,更加直观的反映研究结论,因此该模型具有很强的理论和现实意义;与其他网络估算方法相比,单隐层前馈神经网络极限算法更能符合高速公路建设项目的要求,其预算准确度与广泛应用性较强,在估算流程与速度方面具有明显的优势。利用单隐层前馈神经网络极限算法在计算速度方面的优越性质,给高速公路投资估算提供了一种新的估算方法,具有重要的创新意义。但是,文中亦存在一定的不足。在资料的收集方面,由于应用单隐层前馈神经网络的算法的研究比较少,参考资料不多,这给研究带来一定的难度,可能收集的资料不够全面;经过大量的实践证明,工程特征对其投资的总值具有一定的影响作用,然而在实际预算过程中,难免会遗漏一些影响因素,进而产生误差;在实践证明过程中,与高速公路项目有关的资料、数据、文献、概况等信息(竣工项目、待建项目)是必不可少的,这就加重了信息收集的工作量。所以,获得的案例信息覆盖的地区可能不够广。
[Abstract]:This paper first analyzes the research background and significance, research status at home and abroad, in view of the present status of highway project project investment estimation, to find out the current highway investment estimation problems; introduces the definition of investment of highway and the related investment estimation theory, and the comparative advantages and disadvantages of the existing investment estimation method; at the same time, from the project investment structure, the main characteristics of relevant literature in three aspects, in-depth analysis of parameter characteristics of expressway project investment budget by using grey theory, establishes the investment estimation system of engineering characteristics; grey correlation analysis method and the single hidden layer feedforward neural networks limit combined algorithm, by fully considering the influence factors of expressway investment establishing the model, the 7 parameters introduced highway engineering concepts, and defines it as a model The independent variables, based on the principle of least squares, will directly determine the error within the allowable range, various parameters do not need to repeatedly obtain training quickly calculate, finally the highway engineering investment, scientifically verified by the actual data of this algorithm, in order to highlight the application of good flexibility and adaptability, the research conclusions reflect more intuitive, so the model it has strong theoretical and practical significance; compared with other network estimation method, single hidden layer feedforward neural network algorithm to meet the limit of expressway construction project, the budget accuracy and strong applicability, has obvious advantages in estimating process and speed. Superior properties by using single hidden layer feedforward neural networks limit the algorithm in terms of computational speed, provides a new method to estimate the investment estimate of highway, a new meaning is important but. Yes, it also has some shortcomings. In the process of data gathering, the research and application of single hidden layer feedforward neural network algorithm is relatively small, much less reference, it brings some difficulties to the research, collect data is not comprehensive; after a lot of practice proved that the gross investment of the project features have a certain effect, but in the actual budget process, will inevitably influence factors and some errors; in practice, the verification process, related highway project information, data, literature, survey and other information (the completion of the project, construction project) is essential, this adds to the information collection work. So, in case may obtain the information coverage is not wide enough.

【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F542.3

【参考文献】