Abstract: The logistics distribution networks of business to customer(B2C) e-commerce enterprise are described in a way of an incomplete undigraph, which consists of two kinds of nodes: the distribution center nodes and the customer nodes. A model with 0-1 integer is built to optimize the logistics distribution routing. The model belongs to a type of improved multi-depot vehicle routing problem and has NP-hard complexity. To solve the model, the shortest path and its length of every two nodes in the incomplete undigraph are calculated with FLOYD algorithm, and a predatory search algorithm is designed for the solution. Computation on simulation examples and comparison with genetic algorithm show that the results are satisfactory.