Early prediction of Alzheimer's disease and related Dementia has been a great challenge. Recently, preliminary research has shown that neurological symptoms in Covid-19 patients may accelerate the onset of Alzheimer's disease. With such a further rise in Alzheimer's and related Dementia cases, having an early prediction system becomes vital. Speech can provide a non-invasive diagnostic marker for such neurodegenerative diseases. This work mainly focuses on studying significant temporal speech features extracted directly from the recordings of the Dementia bank dataset and applying Machine Learning algorithms to classify the Alzheimer's disease related Dementia Group and the healthy control group. The result shows that Support Vector Machine outperformed other machine learning algorithms with an accuracy of 87%. Compared to prior research, which used manual transcriptions provided with the dataset, this study used audio recordings from the Dementia bank dataset and an advanced Automatic Speech Recognizer to extract speech features from the audio recordings. Furthermore, this method can be applied to the spoken responses of subjects during a neuropsychological assessment. © 2023, Ismail Saritas. All rights reserved.