Hybrid machine learning approach for anomaly detection

This research aims to <span lang="EN-US">improve anomaly detection performance by developing two variants of hybrid models combining supervised and unsupervised machine learning techniques. Supervised cannot detect new or unseen types anomaly. Hence in variant 1, a model that detects normal samples is followed an screen The weak differentiating between noise fraud. 2, the incorporates validate Three different datasets are used for evaluation. experiment begun with 5 3 models. After evaluation, 2 highest F1-Score one best recall value selected development. 1 recorded across all experiments, indicating it at detecting actual fraud less likely miss compared other can improve precision score significantly original model, better separating from fraud,</span>