Research on the integration of social practice resources in higher education institutions based on deep learning and data mining algorithms

The integration of social practice resources in higher education provides experiential, emphatic, and collaborative learning. Conventional techniques are limited by their inability to involve situational connections and nonlinear interactions between multifaceted datasets, resulting in lower precision. To address these challenges, an intelligent deep learning (DL)-driven framework is developed for efficient integration and optimization of social practice resources with data mining technology. The framework introduces the Attention-based Bidirectional Long-Short Term Memory enriched with Dingo Meta-heuristic Algorithm (Att-BiLSTM-DA), which combines sequential attention learning and meta-heuristic optimization for enhanced analytical performance. It employs an attention-based Bidirectional LSTM for contextual feature representation, while the Dingo algorithm optimizes hyperparameters and selects significant features. A 1000 Social Practice Data in Higher Education dataset pre-processed, using normalisation and noise removal techniques. Moreover, features of textual reflections are extracted and structured participation data using Term Frequency-Inverse Document Frequency (TF-IDF). Att-BiLSTM-DA involves the optimized features for recognizing similarity patterns, aggregating related social practice entities and discovering latent relationships between activities and outcomes. The proposed framework formulates social practice resource integration as a representation learning problem. The attention-based BiLSTM generates contextual embeddings that encode similarity among participation patterns, while the Dingo optimization enhances feature relevance. Classification and regression tasks are used to validate the quality of the learned representations. The implementation was performed using Python and the results show improved accuracy (0.96), and F1-score (0.96) compared to baseline models. The system offers an intelligent, optimized and adaptable approach to handle social practice resources in higher education institutions.