Accurate construction cost calculation is crucial for assessing project viability and selecting design programs. This paper enhances calculation accuracy by first employing the Boruta algorithm to identify vital cost-influencing factors, which serve as the basis for an improved construction cost model. We introduce an enhanced Artificial Neural Network (ANN) model that integrates the AdaBoost algorithm and cost-sensitive methods to refine construction cost estimations. The efficacy of this model is demonstrated through its overall engineering cost error rate of 3.92%, with specific errors in single-side cost, labor, materials, and machinery usage at 3.51%, 7.09%, 3.36%, and 7.93%, respectively. These results meet established accuracy standards, showcasing the model’s potential to significantly improve construction cost management and control.