lzheimer’s disease (AD), the most common form of dementia, is an extremely serious health problem, and one that will worsen in the coming decades as the global population ages. AD is categorized through memory loss, cognitive impairment and inability to do daily routine tasks. This has led to a massive effort to develop both new treatments for the condition and new methods of diagnosis; in fact the two are intimately linked as future treatments will depend on earlier diagnosis, which in turn requires the development of methods that can be used to identify and track the disease. Mild Cognitive Impairment (MCI) is a pre-clinical, pre-symptomatic stage of AD. Patients suffering from MCI have mild memory complaints but intact daily activities. However, only a small percentage of MCI population develop dementia due to AD in the future. MCI patients at risk of developing AD, if identified, monitored and treated, can be given a healthier life. This thesis focuses on early diagnosis of AD by identifying MCI-to-AD progressors. The proposed individual based system takes baseline and one follow-up reading of selected multivariate markers as input and returns the predicted patient status at the subsequent follow-up visit. The process is divided into two phases. The first phase deals with using the available longitudinal data to forecast the marker readings at future time points. For this we propose a novel piecewise linear model with mixed effects from the multivariate markers under consideration. Piecewise linear gradient offsets between subsequent follow-up intervals are modeled as ordinary linear equations modulated by linear prediction coefficients. The linear prediction coefficients are obtained through error minimization and utilized to measure the effect of heterogeneous markers upon the future value of the particular predictor under study. Once the complete time-point trajectory depicting marker flow is obtained, state-ofthe-art supervised machine learning methods are employed to classify the trajectory. These methods include Support Vector Machines (SVM) and Random Forests (RF). This framework is used to identify most effective predictors of conversion in a wrapper based feature selection setup. Furthermore various missing data imputation methods are xi adopted to enlarge the longitudinal dataset size and scrutinize the stability of our tool in presence of artificially generated data. The implemented algorithms are tested and evaluated on publicly available AD dataset using performance parameters such as Mean Absolute Error and Mean Squared Error Area for future value forecasting and Area Under ROC Curve (AUC), accuracy, sensitivity and specificity for classification. The performance improvement of our proposed system is demonstrated by comparing them with recently proposed and published methods