The phenomenon of demand correlation generally exists in the purchase behavior of customers, but most of researches on inventory management assume that product demand is independent of each other, ignoring the impact of correlation and may lead to serious stockouts. In view of the multi-supplier inventory allocation problem with capacity constraint and demand correlation, FP-Growth algorithm was implemented to mine association rules from the transaction dataset where the association rules were abstracted into a kind of constraints. A bi-objective stochastic programming model was built, and piecewise linear approximation method was used to handle the nonlinear objective function and constraint. A non-dominated sorting genetic algorithm (NSGA-Ⅱ) was implemented to solve the model, and improved by the modification of population initialization to accelerate the convergence speed of the algorithm. Numerical experiments using real datasets show that the proposed inventory allocation model based on association rules can significantly reduce the number of stockouts at the expense of slightly increasing the operating cost. Sensitivity analysis on capacity constraint, the confidence levels of association rules and the number of suppliers further provide a reference for the manager of e-commerce platform.