Alzheimer’s disease (AD) is thought to be the most common cause of dementia and it is estimated that only 1-in-4 people with Alzheimer’s are correctly diagnosed in a timely fashion. While no definitive cure is available, when the impairment is still mild the symptoms can be managed and treatment is most effective when it is started before significant downstream damage occurs, i.e., at the stage of mild cognitive impairment (MCI) or even earlier. AD is clinically diagnosed by physical and neurological examination, and through neuropsychological and cognitive tests. There is a need to develop better diagnostic tools, which is what this thesis addresses. Dublin City University School of Nursing and Human Sciences runs a memory clinic, Memory Works where subjects concerned about possible dementia come to seek clarity. Data collected at interview is recorded and one aim of the work in this thesis is to explore the use of machine learning techniques to generate a classifier that can assist in screening new individuals for different stages of AD. However, initial analysis of the features stored in the Memory Works database indicated that there is an insufficient number of instances available (about 120 at this time) to train a machine learning model to accurately predict AD stage on new test cases. The National Azheimers Cordinating Center (NACC) in the U.S collects data from National Institute for Aging (NIA)-funded Alzheimer’s Disease Centers (ADCs) and maintains a large database of standardized clinical and neuropathological research data from these ADCs. NACC data are freely available to researchers and we have been given access to 105,000 records from the NACC. We propose to use this dataset to test the hypothesis that a machine learning classifier can be generated to predict the dementia status for new, previously unseen subjects. We will also, by experiment, establish both the minimum number of instances required and the most important features from assessment interviews, to use for this prediction.