This prospective multicenter study aims to identify the relationship between imbalances in the gut microbiome and abnormal immune activation in individuals with schizophrenia. A total of 297 subjects with schizophrenia (SCZ) and 301 subjects without schizophrenia (HC) provided blood samples for testing. The study also compares those who show immune activation within the SCZ sample (SCZ-A) and those who do not (SCZ-N). After the biological samples were sequenced for gut microbiome composition, all data were run by five machine learning classifiers: A linear classifier (LDA), a probabilistic classifier (Naive Bayes), an ensemble tree-based classifier (GBM), a boosted generalized linear model (glmBoost), and an ensemble decision tree (Random Forest). LDA, Naive Bayes, and GMB showed strong classification between SCZ and HC subjects (AUROC > 0.8), meaning they reliably identified symptoms of schizophrenia using gut microbiome data. For immune activation (SCZ-A vs. SCZ-N), LDA and Naive Bayes performed best (AUROC = 0.927), indicating excellent performance in distinguishing subtypes. In contrast, Random Forest consistently underperformed for both classifications (AUROC < 0.850), suggesting weaker pattern recognition. It is important to note that no external validation was performed. High AUROC values could be due to internal consistency within the small sample sizes rather than due to real-world utility. Furthermore, no temporal validation was established, so the stability of each classifier’s performance over time is unknown. This study's results further understandings about the relationship between the gut microbiome, immune system, and schizophrenia.