This study addresses the preamble detection problem in the Random Access procedure of LTE/5G networks by formulating it as a multi-class classification task and evaluating the effectiveness of machine learning techniques. A Support Vector Machine (SVM) model is implemented and compared against conventional detection methods. The proposed approach improves preamble index assignment, enhancing detection efficiency for User Equipment (UE) accessing the network. Performance analysis demonstrates that the SVM-based solution increases detection accuracy while reducing missed detections. These findings underscore the potential of machine learning in optimizing the Random Access procedure and improving network accessibility.