An improved electrocardiogram (ECG) beats classification system is proposed based on Fuzzy C-Means (FCM) clustering algorithm. The attribute selection model is based on Mahalanobis-Taguchi System (MTS) which can dynamically choose the most relevant ECG features. The concept of initial cluster centroid is applied in order to reduce the number of program iterations of Fuzzy Clustering Techniques. Mahalanobis Distance based FCM (FCM-M) produced the significantly better results as compared to conventional Euclidian distance based FCM in case of Arrhythmia detection.