Design of Engineering Cost Prediction Algorithm Based on Neural Network Model

Abstract

There are major problems in the current project cost management, which is bound to cause huge economic losses to the construction party, the construction party and even the financial institutions. It can be seen that how to reasonably determine the estimate of the early stage of the project to make scientific investment decisions and how to effectively control the project cost is a very important work. Improving the level of project cost prediction is the premise of reasonably determining the project cost, effectively controlling the construction cost, and realizing the lean project cost management, and is the basis of system planning and decision-making. The data in this paper shows that when the number of nodes is small, the acceleration ratio has little difference, and the increasing trend is basically the same. But after the number of nodes increases to 24, the acceleration ratio curve when the data volume is 2000 and 4000 GB has almost no fluctuation, while the acceleration ratio curve when the data volume is 6000 GB continues to show a small upward trend. Through NN (Neural Network), a model can match all existing engineering data samples, and control the error within a certain range. We think this model has a certain prediction function. However, this matching function is not simple linear regression, because any factor affecting the project cost is nonlinear.