A Scalable Artificial Intelligence Framework for Intelligent Data Processing and Decision Support in Modern Computing Systems

Abstract

Today's computing systems are increasingly required to process heterogeneous data and transform it into trustworthy operational knowledge, and artificial intelligence (AI) based decision support is crucial for such systems. For e-commerce applications, user-session data are rich with important behavioral clues, and finding meaningful purchase-intention patterns is difficult due to the mixed data types, class imbalance, and the requirement for interpretable predictive outputs. In this study, a scalable artificial intelligence framework for intelligent data processing and decision support is proposed by using the Online Shoppers Purchasing Intention Dataset. The initial data set consisted of 12,330 records with 18 columns, 17 of which were input features and the last one was the target variable Revenue. Following duplicate removal, 12,205 records were left for analysis. The suggested model included data cleaning, exploratory analysis, preprocessing, model training, model evaluation, and feature-importance interpretation. The data were standardized for continuous variables, categorical variables were encoded, and an 80:20 train-test split was performed on the data in a stratified manner. The following classification models were tested: Logistic Regression, Decision Tree, Random Forest and Gradient Boosting. The accuracy, precision, recall, F1-score, ROC-AUC, Precision-Recall AUC, confusion matrix analysis, training time, and prediction time were used as measures of performance. The results revealed that Random Forest had the highest F1 score which shows the best balance between precision and recall, whereas Gradient Boosting had the highest ROC-AUC and accuracy. The most important predictor was PageValues, as determined by feature-importance analysis. The proposed framework facilitates e-commerce decision making and prediction of purchase intention, customer targeting, campaign planning and conversion optimization.