The fast pace of the evolution of smart education platforms, it is possible to leverage the interaction data of the learner to create more advanced personalized learning analytics. Current approaches to prediction of student performance and the adaptivity of learning systems are not optimal in feature selection, have poor prediction accuracy, are not very scalable, and do not effectively deal with heterogeneous educational data. In the modern digital learning system, such constraints reduce the effectiveness of smart tutoring systems and personalised recommendation systems. The paper suggests an Intelligent Pattern Computation framework based on a fusion of Deep Learning (DL) and Swarm Optimization (SO) to address these challenges and personalize learning analytics for intelligent education systems. The proposed model employs Deep Neural Network (DNN) to find out higher order behavioral and academic learning and Particle Swarm Optimization (PSO) for feature selection and hyperparameter tuning. The hybrid system improves the predictions, learner profiling and the effectiveness of adaptive content recommendations. The benefit of the proposed framework is that it can be predicted with high precision at a lower level of computation at a given level, but with a higher level of flexibility in meeting various types of learner behaviours. The model is evaluated on benchmark education data sets such as student performance data in UCI Machine Learning Repository and data sets of interaction of online learning examples. Experimental results have shown that it performs better than existing machine learning techniques with 98.2% accuracy, 97.4% precision, 96.9% recall, 97.1% F1-score and 2.3% lower error rate. The proposed system plays a significant role in personalized learning, drop-out prediction of students in early stages and smart academic decision making. It enables sustainable, data-informed, intelligent learning ecosystems with effective adaptation of learners, enhanced engagement and optimized learning outcomes.