Study On Cost Prediction Model About Road Engineering Of Municipal Maintenance Project

With the improvement of urban infrastructure construction in China,the tasks of maintenance management are becoming increasingly serious.It is urgent to initiate a market-oriented reform of separating management from maintenance in municipal road and bridge maintenance industry.The lack of scientific valuation standards is becoming a bottleneck of market-oriented reform in municipal road and bridge maintenance industry.Therefore,the exploration of scientific valuation standards for municipal maintenance project plays an important role in promoting the reform of municipal maintenance industry.In this article,by comparing municipal maintenance projects conditions with list valuation quota arrangement conditions,the reasons why actual cost of the municipal maintenance projects differs from current valuation standards price is found out.First,the maintenance projects have single content that involves few listing items,small and scattered quantity.Second,the maintenance project is mostly affected by human..Usually,there are high standards in the safety requirements in municipal maintenance projects.That is to say,the actual situation of municipal maintenance projects is very different from the preparation condition of list valuation quota.Therefore,the quantified indexes of different conditions on municipal maintenance project and list valuation are listed,which entail quantity,number of maintenance positions,night working duration,waiting duration and road types.The view that the real cost of the municipal maintenance projects is influenced by the current list valuation quota set prices and factors listed above is put forward.The regression analysis confirmed that the independent variables were linearly correlated with the dependent variables after logarithmic or combinatorial transformation,and the parameter R2 of regression model is 0.921.The artificial neural network is established by adopting BP algorithm and RBF algorithm respectively.The dependent variables and independent variables were fitted,and the parameter R2 of neural network model is above 0.9.The optional parameters lies in that the average Ei is 16.23%,MAD being 1.75 and MSE being 4.3.Compared with the predicted values of regression,BP and RBF neural networks respectively,the RBF neural network predicted the best results in the aspect of predicting average Ei,up to 16.23%.On the basis of absolute deviation of MAD and mean square error MSE,BP algorithm has the optimal neural network prediction effects,with 1.75 and 4.3 respectively.Furthermore,the gap of BP is the lowest in comparison with the actual values and prediction ones of ten test sets.Finally,on the basis of inadequacy of this theoretical research,the research direction and methodology of the next phase is put forward.This research provides a solid foundation for establishing scientific and authoritative valuation standards of municipal maintenance industry.