Comparative Performance of Random Forest Algorithms and Discriminant Analysis in Classifications

The use of computer-based systems plays a leading role today as an analytical technique in diagnosing diseases, including predicting autism disorders. This research aimed to predict the class or target by using related variables accurately. The prediction process involved the number of attributes and records from the dataset. The dataset used in this study was the autism disorder dataset taken from the UCI repository machine learning; the data was already in number (binary) thus, no dummy variable was necessary is not needed. In this study, we used the Random Forest algorithm and discriminant analysis. The purpose of this study was to predict the type of autism disorder in children according to their prescribed symptoms. Two types of disorders used as target variable; communication disorders and behavioral disorders. Meanwhile, the independent variables are symptom 1, symptom 2, symptom 3, symptom 4, and symptom 5. The results from these are later processed using the SPSS software. The results from the Random Forest method indicated that 74.2% of forming a classification and a sensitivity value of 26.3%. Meanwhile, the percentage in the discriminant analysis was 70.1%. This value indicated that the decision tree was more reliable in classifying target types