Due to the rising complexity of software projects, it is quite difficult to predict the risk in software requirements which is the most profound and essential activity in SDLC. It may lead to the failure of a software project. Risk prediction in software requirements is more crucial as it is the start of any software project. In this study, we propose an ensemble classifier based on AdaBooostM1 and J48 combinedly named as (ABMJ), for risk prediction in software requirements. The performance of the proposed ABMJ is compared with seven diverse ML algorithms including A1DE, MLP, CSF, J48, NB, RF, and SVM. These ML models are evaluated on the risk dataset available at Zenodo repository based on the accuracy, MCC, F-measure, recall, and precision. The overall analysis shows the best performance of ABMJ with an accuracy of 97.6285% and the worst performance of MLP with an accuracy of 62.0553%. This study's analysis may be used as a standard for other academic studies, allowing the outcomes of any proposed approach, framework, or model to be benchmarked and essentially established.