This study aims to develop a novel biomarker based on Machine Learning (ML) classifiers to predict the severity of Autoimmune Thrombocytopenic (AITP) bleeding. Traditional clinical notification approaches fail to forecast AITP patients’ conditions, resulting in inappropriate treatment and patient risks. ML classifiers, such as Adaptive Genetic Algorithm (AGA), Support Vector Machine (SVM), and Naïve Bayes (NB), classify diseases; however, a novel biomarker based on ML classifiers faces a gap in research. Two publicly available Gene Expression Omnibus (GEO) datasets were used to address this. Three efficient ML models were used to normalize data to better predict acute and chronic AITP after a significant thrombocytopenic event. The findings show that, the proposed AGA model outperformed the SVM and NB models with 97.5% accuracy, 95.2% sensitivity, 100% specificity, and 98.7% F1 score and hence, it is expected that this ML-based biomarker should help doctors and lab technicians in anticipating acute and chronic AITP.