Sequential Pattern Mining with Machine Learning

With the exponential rise in digital data, extracting meaningful patterns from large-scale sequential datasets has become a pressing need across domains such as healthcare, retail, and cybersecurity. This study explores an integrated approach combining Sequential Pattern Mining (SPM) and machine learning to identify both positive and negative associations in sequential data. Using the Instacart Online Grocery Shopping Dataset, frequent item sets are extracted and categorized through association analysis. A hybrid classification model incorporating Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and Support Vector Machines (SVM) are proposed to enhance predictive accuracy. The model captures temporal, spatial, and discriminative features, achieving superior performance by achieving an accuracy of 96.5%, precision of 98.63%, recall of 90.0%, and F1-score of 94.0%, significantly outperforming standalone models like CNN (70.0%), Random Forest (92.2%), and ensemble methods (88.84%). Additionally, the Area Under the Curve (AUC) for the hybrid model reached 0.80, compared to 0.54 for SVM and 0.57 for CNN. These results underscore the potential of combining pattern mining with deep learning for robust and efficient sequence classification and decision-making in real-world datasets.