Objectives
​
The main goal of this dissertation is to assess how useful Data Mining methods and algorithms can predict the adverse effect of drugs.
In the first stage, the main objectives are not only study and understand the basic concepts inherent to the adverse effects of drugs, but also analyze sites and databases where data and information concerning ADRs is available. After this research, it is important to investigate studies and methods that already exist to predict these effects and, finally, explore the tools of Data Mining and assimilate the interest of this method in predicting adverse effects of drugs.
The idea is to choose drugs, whose effects are known and available in a reliable database, and use a Data Mining tool to predict such known effects. That is, in order to verify if the tool used is feasible to predict unknown adverse effects of drugs, is performed a simulation where if supposes that adverse effects of selected drugs are not known and, through the crossing of data of other drugs, it is intended discover the adverse effects of drugs.
In this work two experiments, involving Recommendation Systems (RSs) and Classification algorithm were performed. RSs have as their main objective to filter information and provide researchers with only relevant and highly correlated information. The recommendation algorithms used were Matrix Factorization (MF), Slope-One and User k-NN. Commonly, DM has the ability to perform, among others, classification tasks. The Classification is known as "the process of learning a function that maps (classifies) a given object of interest into one of the possible classes". ​We have also defined a classification task and have used Decision Tree, Random Forest, Naive Bayes and Support Vector Machines (SVM).
​
