Insurance companies leverage machine learning capabilities, deep domain expertise and technology competency of Wipro HOLMESTM to combat fraudsters. HOLMES identifies new and previously unknown fraud schemes with high accuracy by analyzing and understanding patterns. Special investigation units focus on actual red flags as the false positive rate reduces through continuous learning, feedback and self-calibration over time. Only a small set of identified frauds are required to train HOLMES. Separate models can be built for each fraud type such as staged theft, staged fire etc. Predictions also provide indicators for flagging a case, such as suspicious witness and incident during the holiday period.