A Hybrid LSTM and Decision Tree Model: A Novel Machine Learning Architecture for Complex Data Classification

We propose a novel hybrid machine learning architecture that synergistically combines the temporal data processing capabilities of Long Short-Term Memory (LSTM) networks with the categorical data handling and interpretability features of decision trees. This architecture addresses the complex problem of classifying mixed sequential and non-sequential data, a common challenge in various computer science domains. Our approach harnesses the strength of LSTM networks in processing sequential patterns and the power of decision trees in handling categorical data. To evaluate the effectiveness of the hybrid model, we conducted rigorous experiments using a two-year dataset from a university. Benchmark tests against standalone LSTM and decision tree models showed that our hybrid architecture consistently outperforms in terms of accuracy, precision, recall, and F1 score. The results attest to the potential of hybrid models in managing complex data classification tasks, thus making a significant contribution to the growing field of machine learning model fusion.