Wireless data broadcast system can support data access services for any number of the clients. In the system, it is critical for the performance in order for the server to disseminate the preferred data items of the clients to the wireless channel. This paper proposes a query likelihood model that can predict data preference of the clients in near future. The proposed model predicts the query likelihood by considering the data preference of the clients and the cell preference in a hybrid manner. The cell preference means the probability for the clients to reside within the broadcasting service area. The data preference is modeled with an artificial neural network in linear regression. With the proposed query likelihood prediction model, the broadcast server enables to reflect actively the needs for data items of the clients in the near future. Through intensive simulations, the effectiveness of the proposed model is shown with respect to the access time and tuning time of the clients in the broadcasting system.