The mobile phone has become a necessary tool in daily life, thus generating bulk call and message logs. The increased use of telecom services also attracts the attention of fraudsters to misuse the network or service. This article addresses the identification of forged instances where consumers are tricked into paying for services they are unaware of while on the phone. A novel approach is designed using the ensemble classifier to detect fraudulent activities in the user call records. The few numbered malicious examples are initially subjected to an oversampling technique to produce more synthetic minor class data. Later, an ensemble learner processes the balanced call data and segregates the genuine users from intruders. The performance of the model is assessed by utilizing real-world call data. This work additionally presents the effect of imbalanced data on the classifiers when identifying telecom fraud.