Screening important indices based on the historical construction of a distribution network to make targeted and precise investment decisions is an effective way of improving the effectiveness of network construction. This paper uses multi-source fusion technology to integrate the construction data and run data of different systems in the distribution network, and uses textual data extraction technology to structurally transform the project construction record. The extracted data can be screened based on an Elastic-net and the parameters of the Elastic-net are optimized based on the BIC criterion. In this way, the importance ranking of construction indices is obtained. From this ranking a distribution network construction investment decision-making model is established. The comparative analysis of different investment decision-making schemes shows that the construction effect of the proposed model is more pertinent, and can achieve precise investment in the construction of distribution network projects.