<![CDATA[Mental health disorders are often difficult to detect and diagnose, causing misdiagnoses which lead to inappropriate treatment and have a negative impact on the sufferer's quality of life. This research aims to develop an accurate and efficient model for identifying mental health disorders by utilizing the Random Forest method and Cross Validation techniques. Random Forest was chosen because of its ability to improve prediction accuracy and training speed. Cross Validation is used to train and test models with various combinations of data, and reduces the risk of Overfitting. The dataset consists of 120 data with 18 behavioral attributes and diagnoses, with four target classes: Bipolar Type-1, Bipolar Type-2, Depression, and Normal. Four Cross Validation experimental scenarios were tested: k=5 and k=10, and k=5 and k=10 with Stratified to reduce data bias. Experimental results show that k=10 stratified cross-validation produces the highest accuracy (87.5%), with precision, recall, and F1-score also reaching 87.5%. The Stratified technique is proven to improve the balance of class distribution and reduce the risk of Overfitting. These findings confirm that Random Forest with k=10 Stratified Cross-Validation is the optimal approach for diagnosing mental health disorders. The implications of this research include the potential for applying models in AI-based systems to assist medical personnel in more accurate and efficient early diagnosis.]]>
Copyrights © 2025
