Depression is a mental disorder that might cause self-harm or suicide. Selective serotonin reuptake inhibitors (SSRI) therapy is the most prescribed medication to treat depression. This paper proposes a novel classification model to predict the outcomes of SSRI therapy for depressed patients based on recorded electroencephalogram (EEG) signals. Our proposed method is developed using a tunable Q-factor wavelet transform (TQWT) as a robust signals processing technique. The EEG signals are decomposed using TQWT into six sub-bands, and the log energy entropy is computed as a feature for each sub-band. The statistically significant features are selected by the Kruskal-Wallis test. The selected features are used as inputs for k-nearest neighbors (KNN) classifier. Mumtaz database is used to evaluate the performance of our proposed method. Mumtaz database has pretreatment EEG signals of 30 major depressive disorder patients who were selected to start SSRI therapy. The results show that the proposed method achieves a promising classification accuracy level 99.17% using 10-fold cross-validation strategy. The results show that our method outperforms other methods in the literature. The results show the importance and suitability of the higher frequency spectrum of EEG signals to predict the outcomes of SSRI therapy. The proposed method can be used in clinics and hospitals as a computer-assisted prediction system.