Neural network based composite project risk evaluation model

Abstract

The invention relates to a neural network based composite project risk evaluation model. As can be shown in the figure, the model is expanded, on the basis of a traditional AHP method to set weight structure, from the third-level index to a three-type composition structure featuring a neural network layer, a quantization adjustment layer and a filling layer. The filling layer and the quantization adjustment layer are capable of importing the information filled by the project party for initial combined quantization to form the input vector of the neural network layer; then, after second quantization calculation through the neural network layer, the third-level index layer input value of a traditional risk evaluation model is developed; finally, through the weight calculation of the traditional three-level indexes, a final risk evaluation result is obtained. The neural network layer, after some data trainings, is able to replace the evaluation person to give scores not so subjectively to some extent so as to achieve evaluation through artificial intelligence. At the same time, all a project party needs to do is input the unitized basic information which is gathered and combined by a hidden quantization adjustment layer and is used as input vector of the neural network layer. This prevents the false or exaggerated information from the project party and makes the result more authentic.

Description

[0001] Technical field [0002] The present invention relates to a new type of project risk assessment model. On the basis of the traditional three-level evaluation index model based on the AHP method to set the weight, the model expands the neural network layer, quantitative adjustment layer, filling layer, etc. from below the third-level index. Three types of composite structures. The filling layer and quantification adjustment layer can import the information reported by the project party and perform primary combination quantification to form the input vector of the neural network layer, and then pass through the secondary quantitative calculation of the neural network layer to form the third-level index input of the traditional risk assessment model Finally, the final risk rating result of the project can be obtained through traditional three-level indicators and weights. After a certain amount of data training, the neural network layer can replace the subjective scoring pr...

AI Technical Summary

Problems solved by technology

[0004] In order to overcome the traditional risk assessment model’s weak quantitative analysis, strong artificial subjective factors, and inability to adapt to complex time-varying environments, this invention introduces a new type of project risk assessment model, considering the labor cost of the classic AHP model relative to the Delphi model Less, the model of the present invention continues to use the traditional AHP method as the basic model for determining the index weights at all levels, and at the same time, expands three types of models such as the neural network layer, the quantitative adjustment layer, and the filling layer from below the third-level index to form a composite structure. The primary quantification of the evaluation input information is realized through the quantitative adjustment layer, and the neural network model is used as the evaluation algorithm of the third-level index with the largest number, replacing the manual evaluation process of the Delphi model

Images