The joint optimization of slot allocation and empty container repositioning is a critical issue in the liner shipping industry. To address the uncertainty in transportation demand, this study first develops a data-driven forecasting approach using a hybrid LSTM-MLP model, trained on historical big data to predict segmented market demand. The forecast model combines a long short-term memory network (LSTM) and a deep multi-layer perceptron (MLP). Based on the predirection, a multi-period joint optimization model is proposed that integrates slot allocation and empty container repositioning under a diversity service strategy. A mixed-integer programming formulation is developed, and a novel branch-and-cut algorithm is designed to enhance solution efficiency. Results demonstrate that the joint optimization model based on the diversity service strategy can effectively improve the revenue of liner companies and increase customer satisfaction. Four groups of experiments verify the effectiveness and accuracy of the branch-and-cut algorithm. The proposed joint optimization results based on LSTM-MLP prediction can increase the total revenue by 8% $\sim $ 17% compared with the joint optimization under random scenarios.