Ventricular fibrillation is an irregular heart rhythm that can occur if there is a cardiac attack or other heart muscle injury. It manifests through rapid and irregular contractions of the lower chambers of the heart, causing heart failure and impairing the efficient circulation of blood throughout the body. Since ventricular fibrillation (VF) is a potentially fatal illness that necessitates quick medical attention and its accurate diagnosis is crucial for cardiologists. Several research studies have been conducted that show promise in utilizing conventional heart rate variability parameters to detect VT and VF. One type of approach obtained 92.43% average accuracy which used DWT based feature extraction and Support Vector Machine for VF detection. Hence, further research is needed to enhance detection accuracy as VF can be fatal. To address this need, an approach has been proposed to detect ventricular fibrillation.In this approach, the significant features are extracted from the MIT-BIH database using Principal Component Analysis (PCA). The MIT-BIH arrhythmia database includes around 1715 patient’s normal and ventricular fibrillation ECG signals. Support Vector Machine(SVM), Logistic regression(LG),Gaussian Naivy Bayes(GNB), Decision tree(DT) and Random Forest(RF), Adaboost(AB) and LSTM models were trained and tested with 1372 and 343 samples. Accuracy, precision, specificity, recall, Fscore, receiver operating characteristic curve (ROC) and sensitivity have been measured to evaluate the prediction models. Among all models, SVM demonstrated better and more convincing performance for this specific dataset. It employs a nonlinear kernel, enabling it to approximate any computable function and effectively separate various types of datasets. Therefore, the SVM prediction model holds promise in enhancing performance in VF detection.