This paper presents a student performance prediction model (MTAPSP) that leverages multidimensional time-series data analysis to predict student performance in online learning environments. The model integrates students' learning behaviors, assessment scores, and demographic information to capture the complex relationships between these factors and predict student outcomes. The MTAPSP model consists of two main components: a multi-layer LSTM network with a multi-head self-attention mechanism (MHSA) and an artificial neural network (ANN). The MHSA mechanism enhances the model's ability to extract and weight the importance of temporal behavioral features, assessment scores, and demographic features. The ANN layer integrates these features to improve the model's performance. The model was evaluated using the Open University Learning Analytics Dataset (OULAD), achieving 74% accuracy and 73% F1 scores in multi-classification prediction tasks and 99.08% accuracy and F1 scores in early risk prediction tasks. Compared to benchmark models, the MTAPSP model demonstrated superior performance in both multi-classification and early prediction tasks. The study highlights the importance of considering multiple dimensions of student data and the effectiveness of the proposed model in providing timely interventions to support at-risk students and enhance overall learning outcomes.This paper presents a student performance prediction model (MTAPSP) that leverages multidimensional time-series data analysis to predict student performance in online learning environments. The model integrates students' learning behaviors, assessment scores, and demographic information to capture the complex relationships between these factors and predict student outcomes. The MTAPSP model consists of two main components: a multi-layer LSTM network with a multi-head self-attention mechanism (MHSA) and an artificial neural network (ANN). The MHSA mechanism enhances the model's ability to extract and weight the importance of temporal behavioral features, assessment scores, and demographic features. The ANN layer integrates these features to improve the model's performance. The model was evaluated using the Open University Learning Analytics Dataset (OULAD), achieving 74% accuracy and 73% F1 scores in multi-classification prediction tasks and 99.08% accuracy and F1 scores in early risk prediction tasks. Compared to benchmark models, the MTAPSP model demonstrated superior performance in both multi-classification and early prediction tasks. The study highlights the importance of considering multiple dimensions of student data and the effectiveness of the proposed model in providing timely interventions to support at-risk students and enhance overall learning outcomes.