In a cybersecurity operation center, an organization is implementing an AI-driven anomaly detection system to enhance its threat detection capabilities. The system is designed to analyze network traffic patterns and identify deviations from established baselines. During the initial phase, the AI model is trained using historical data, which includes both benign and malicious traffic. After deployment, the system flags a significant number of alerts, but upon investigation, many of these alerts are false positives. What is the most effective approach to improve the accuracy of the AI model in this scenario?