Bipolar disorder is a common and serious mental illness affecting millions worldwide. Severe mood swings, such as bouts of sorrow or mania, may significantly influence a person's daily functioning and quality of life. This study proposes an effective technique for predicting bipolar disorder by combining complex data preprocessing, feature selection, clustering. The approach includes an upgraded decision tree and a weighted k-means approach for missing data preparation and analysis. Relevant feature selections are made using a weighted binary BAT approach, reducing the data's dimensionality while preserving essential qualities. The data is then clustered using a hybrid approach combining the particle swarm optimization (PSO) algorithm with the Gaussian Kernel fuzzy clustering algorithm (GKFCM), resulting in a clustering accuracy of 98.6%. Overall, the proposed approach offers an effective and efficient way to predict bipolar disorder with promising results, and it has the potential to be extended to other related illnesses