Desarrollo de modelos de aprendizaje automático aplicados a la patología de Alzheimer : Modelo de aprendizaje automático no-supervisado para el análisis de pacientes con deterioro cognitivo relacionado con la patología de Alzheimer.

Abstract

The progress of Alzheimer's patients has been chosen as the topic for this paper to curricular integration, in order to define the development of an unsupervised learning model, which was carried out from a set of data. The development of the model was worked in the Google Colab work environment, which showed the need to use a treatment of missing values on the data, in order to carry out the subsequent execution of the corresponding algorithms. Then the optimization was carried out between the variables, in order to eliminate those that cause noise within the analysis. From this, the data were scaled to perform principal component analysis (PCA) and reduce the dimensionality of the data set and thus be able to observe the separability of the data. The application of clustering allowed Alzheimer's patients to be assigned to the status group to which they belong, that is, dementia, non-dementia, and transition or conversion. Unsupervised learning algorithms are not susceptible to measurement in percentage of certainty, the results obtained from this work can be taken as a basis for supervised learning algorithms to allow prediction of the progression or not of the disease towards a diagnosed dementia.