【作者】 张媛媛；

【导师】 陈小波；

【作者基本信息】 东北财经大学 ， 工程管理， 2020， 硕士

【摘要】 投资估算是项目前期决策的关键环节,对于控制项目的成本十分重要,能够决定一个项目的成功与否,工程实践证明,前期的投资决策对项目成功的影响程度高达70%。因此,建立一套科学准确有效的估算方法是至关重要的。传统的投资估算主要依靠估价师的个人经验,采用拟建工程与已建工程的相似程度进行类比得出拟建工程项目的估算造价,有生产能力指数法、系数估算法等,这些估算方法速度快,但准确度不高。随着计算机技术的进步,越来越多的学者开始研究如何使用人工智能估算方法来提高投资估算的准确性与科学性。其中,人工神经网络、支持向量机等方法由于在解决非线性回归问题上具有很大优势,在许多研究中都得到充分应用。此外,为了提高人工智能方法的估算精度,更多研究者们也将优化算法引入估算过程中,例如遗传算法、粒子群算法、差分进化算法等。前人研究在很大程度上完善了人工智能估算体系,弥补了传统估算中估价师根据自身经验进行估算的缺陷,使得估算精度有了很大提高。论文主要对建筑工程项目投资估算的科学性与准确性作出研究,不同于现有研究,本文创新性地将估算过程中存在的不确定性和风险考虑在内,并以概率性置信区间的形式体现出来。如此,得到一种更加切实可行的人工智能估算方法,用来辅助前期投资决策,提高决策效率。在本研究中,提出一种灰狼优化算法,通过引入灰狼优化算法来解决在投资估算时只考虑点值,缺乏理性考虑,忽视预测中存在不确定性的问题,提出一种概率置信区间预测方法,大大降低了决策风险。针对现有的研究在投资估算预测值的确定时仅取点值,缺乏对于估算不确定性的合理考虑,使得决策风险大幅增加的问题,论文引入灰狼优化算法提高点预测模型的预测精度,并在此基础上提出一种概率性置信区间预测方法。首先,在对工程项目案例有一定搜集之后,本文采用SPSS统计分析软件的探索性因子分析功能对初步选取的17个工程特征指标降维综合提取公共预测因子,这些指标既包含定性指标也包含定量指标,其中定性指标在降维前已得到量化。在探索性因子分析之后,所提取出的6个公共因子是相互独立的,并且可以代表原有指标的绝大部分信息,它们将作为模型的输入变量。然后根据52个实际案例经过线性插值形成的152个样本建立支持向量机回归模型,随机划分122个训练集样本和30个测试集样本,使用训练集样本训练模型,使用测试集样本测试模型的性能。在模型训练过程中,采用灰狼优化算法寻找支持向量机回归模型的最优参数,提高点预测模型的预测精度。最后,根据刚才得到的点预测模型的预测误差,创新性地提出一种基于预测误差分布特性统计分析的非参数置信区间估计方法,主要通过核密度估计方法求得预测误差的概率密度函数,并将概率密度函数转化为概率分布函数,从而求得不同置信度下的投资估算概率性预测区间。论文通过多组案例研究结果表明:在经过灰狼算法优化后,本文的支持向量机回归模型的预测精度高达93%;在设置置信度为95%时,区间预测的综合评价指标CWC为2.17,成本估算的区间覆盖率PICP为93.33%,这说明确定性预测结果和概率性预测结果都能得到符合预期的效果。因此,本文提出的区间预测方法可靠性较高,模型具有有效性,进一步完善了现有的人工智能估算体系,可以对工程建设项目前期的投资决策做出具有现实意义的指导,并为决策人员提供更加丰富的预测信息,减少前期投资决策的不确定性,增加预测模型的抗风险能力,对项目成功具有十分重大的意义。更多还原

【Abstract】 Investment estimation is a key link in the early stage decision making of a project,which is very important for controlling the cost of a project.It can determine whether a project succeeds or not.Engineering practice has proved that the early stage investment decision has an impact on the success of the project up to 70%.Therefore,it is very important to establish a set of scientific,accurate and effective estimation methods.The traditional investment estimation mainly relies on the personal experience of the valuer,and USES the similarity degree between the proposed project and the already built project to draw the estimated cost of the proposed project,including the production capacity index method and coefficient estimation method,etc.These estimation methods are fast but not accurate.With the progress of computer technology,more and more scholars begin to study how to use artificial intelligence estimation method to improve the accuracy and scientific nature of investment estimation.Among them,artificial neural network,support vector machine and other methods have great advantages in solving nonlinear regression problems and have been fully applied in many researches.In addition,in order to improve the estimation accuracy of artificial intelligence methods,more researchers have also introduced optimization algorithms into the estimation process,such as genetic algorithm,particle swarm optimization,differential evolution algorithm,etc.Previous studies have improved the artificial intelligence estimation system to a large extent,and made up for the defects of the traditional estimators’ estimation based on their own experience,which greatly improved the estimation accuracy.This paper mainly studies the scientificity and accuracy of the investment estimation of construction projects.Different from the existing research,this paper innovatively takes the uncertainty and risk existing in the estimation process into account,and presents it in the form of probability confidence interval.Thus,a more practical and feasible artificial intelligence estimation method is obtained to assist the early-stage investment decision and improve the decision-making efficiency.Aiming at the problem that the existing research only takes the point value when the forecast value of investment estimation is indeed fixed,and the decision risk is greatly increased due to the lack of reasonable consideration of the uncertainty of estimation,this paper introduces the grey Wolf optimization algorithm to improve the prediction accuracy of the forecast model of high point,and on this basis,proposes a probabilistic confidence interval prediction method.First,after gathering in the project case to a certain extent,this article adopts the exploratory factor analysis of SPSS statistical analysis software features a measure of the preliminary selection of 17 engineering characteristics of dimension reduction comprehensive extract public predictors,the index contains both qualitative index also includes quantitative indicators,including qualitative indicators before dimension reduction has been quantified.After exploratory factor analysis,the six common factors extracted are independent of each other and can represent most of the information of the original indicators,and they will be used as the input variables of the model.Then the regression model of SUPPORT vector machine was established according to 152 samples formed by linear interpolation in 52 actual cases,and 122 training samples and 30 test samples were randomly divided.The training model and the performance of the model were tested using the training model.During the model training,the grey Wolf optimization algorithm is used to find the optimal parameters of the SVM regression model and improve the prediction accuracy of the forecast model.Finally,according to just get a point prediction model prediction error,innovative put forward a kind of based on statistical analysis of the prediction error distribution characteristics of confidence interval estimation method,mainly by kernel density estimation method for the prediction error,the probability density function and probability distribution function and probability density function can be converted to investment estimation under different confidence level obtained and probabilistic prediction intervals.The results of multiple case studies show that the prediction accuracy of the SVM regression model is as high as 93%after the grey Wolf algorithm is optimized.When the confidence is set as 95%,the comprehensive evaluation index CWC of the interval prediction is 2.17,and the interval coverage PICP of the cost estimation is 93.33%,which indicates that both the deterministic prediction results and the probabilistic prediction results can achieve the expected effect.Therefore,this article puts forward the reliability interval prediction method is higher,the model is effective,further improve the existing system of artificial intelligence estimate for engineering construction projects of investment decision making has the realistic meaning of guidance,and provide decision makers with more abundant forecast information,reduce the uncertainty of investment decision-making,increase the ability to resist risk prediction model,it is very significant to project success.更多还原