Abstract

Objective: To develop a reliable model for the early diagnosis of schizophrenia (SZ).
Methodology: Using a schizophrenia dataset from Kaggle, linguistic values replaced numerical gradations to better reflect symptom categories (e.g., "yes"/"no", "high"/"low"), an automated system-cognitive analysis was performed on a dataset of 9,238 patients aged 18-80, with 2,803 diagnosed with SZ and 6,435 not diagnosed. The Eidos intelligent system developed statistical and system-cognitive models based on the training data.
Results: Three statistical and seven system-cognitive models were generated. These models formed information portraits of SZ by identifying symptom sets and their influence on disease presence or absence. The system enabled the identification of the most significant symptoms, simplifying diagnosis and improving predictive accuracy. A SWOT analysis further classified symptoms into those that promote or inhibit disease progression and their relative impact.
Conclusion: System-cognitive modelling enhanced early detection of SZ, helping reduce diagnostic effort and improve treatment outcomes. Early intervention can prevent severe disability and lower healthcare costs.

Key words: Schizophrenia, positive and negative symptoms, GAF scores, Kaggle repository, genetic predisposition.