Projects of engineering construction have the characteristics of large investment and long cycle, which makes the cost management difficult and the data are often abnormal. Therefore, it is necessary to strengthen the detection of abnormal data in engineering cost list. Based on this, the establishment of a detection model of engineering cost list is studied in this paper. By introducing K-means clustering method into the model, the list is clustered according to the comprehensive unit cost, and the list data are classified by Bayesian list classification method where the value of k is selected as 5. The detection of abnormal data method in engineering cost list is compared with that of the traditional detection method based on distance, which is known that the detection model has good effect, high accuracy and recall rate.