In this study, Electrodermal Activity (EDA) signals were analyzed to evaluate the changes between physical stress, cognitive stress, and emotional stress. For this purpose, energy and variance properties of the EDA signals in the time domain were analyzed for each case and as short-time frames. In addition, the EDA signals were decomposed using the Empirical Mode Decomposition (EMD) method, and the sub-band signals were analyzed for each case. Further, the Short Time Fourier Transform (STFT) method was used to analyze the in the time-frequency domain of these signals. Also, according to obtained features, EDA signals were classified to determine the stages. Simulated results show that, the EDA and subband EDA signals were found to be significantly different in terms of cognitive stress (p<0.05). Also, the features obtained from the EMD subbands were classified using Support Vector Machines (SVM), K-Nearest Neighbors (KNN), and Multi-Layer Perceptron (MLP) methods for different situations and classifier performances were compared. In the classification of cognitive stress period and first rest period, the best classification performance was achieved as 97.36 %, 84,21 %, and 81,57 % using MLP, SVM and KNN classifier respectively compared to other situations.