Many times at different situations like meetings and phone calls, knowing the speaker's identity is useful. Speaker diarization, which segments and categorizes a voice signal to the speaker identification, can be used to do this task. The i-vectors are taken out of the speech segments and used to do speaker diarization. Using both supervised and unsupervised machine learning algorithms, a model is developed for the retrieved attributes. The fresh voice segments from the speakers can be categorized using this trained model according to the corresponding speakers. Both supervised and unsupervised speaker diarization will be carried out in this study of review paper. The analysis of voice files is extremely relevant and significant in today's world of abundant voice files. In this study, the authors use supervised and unsupervised machine learning methods to propose a speaker diarization. Data processing, extraction and classification, data segmentation, and the learning phase are the system's four major segments. SVM and Multilayered- Neural Networks, two standard supervised, & the unsupervised learning technique of k-means clustering were used to train and validate the model. The authors also give an ensemble of the characteristics, which would be determined to perform much better overall, according to their findings