Improving the Quality of Enterprise Data Management with Tree-Based Models

Data credibility is essential for reliable decision-making and decision-support systems across Salesforce environments during the processing of transactional data as observed in the Online Retail Dataset. This work analyses the application of tree-based machine-learning models in improving data quality through transformation and cleaning processes for the removal of missing values, duplication, and outliers. The approach includes data preparation, relevant feature selection, model construction, and deployment within Salesforce processes to monitor data quality in real time and batch workflows. With tree-based models, substantial performance gains were observed, with precision up to 90.8%, recall up to 74%, and overall accuracy up to 91.6%. After Salesforce integration, completeness increased by 12%, accuracy by 10%, and consistency by 15%. The system’s retraining mechanism and feedback loop ensure protection against long-term data degradation in enterprise CRM environments.