SC4ANM: Identifying Optimal Section Combinations for Automated Novelty Prediction in Academic Papers