Online customer reviews are invaluable tools for making informed decisions, yet this utility is affected by generation of Fake Reviews. This problem equally concerns both consumer and business fields. Existing Methods are based on linguistic attributes. With the incoming of various open-source LLMs, the distinction between machine and human generated text is fading. In this research, we observed both linguistic and behaviour-based detection systems provide better accuracy. We adopted a hybrid approach using Semantic Analysis and supervised ML approach on Yelp Review dataset. Various useful attributes were used/formulated to form new parameters. Semantic Analysis has been used to form certain attributes in the dataset to include the linguistic behaviour of review. Our study is grounded on comparative analysis of 10 different supervised ML models. Comprehensive evaluation of models is performed using various metrics. The best achieved accuracy was 94.362% with ROC AUC of 0.982, precision of 0.915, recall of 0.978 and F1 Score of 0.945. The detailed experimental analysis shows the feasibility and effectiveness of proposed system.