This paper addresses an important problem in the management of daily synchronous line loss rate (SLLR): how to quantitatively assess the effects of each influential factor and identify the most critical ones? An SLLR calculation method based on the K-modes clustering algorithm and the deep belief network (DBN) is proposed. First, the influential factors of the SLLR in the transformer district are determined. Then, the transformer districts are classified into different clusters by using K-modes algorithm to balance the accuracy and computation burden. Next, the DBN based SLLR calculation model is built and trained for each cluster of transformer districts. In the end, the sensitivity analysis is conducted to quantify the effect of each factor and identify the most influential ones. Historical data from the power supply company is used as test case to verify the effectiveness and correctness of the proposed methodology.