Aim:
Aims: to identify clinical and Sociodemographic predictors of anxiety and depression among cancer patients undergoing active treatment using XGBoost algorithm.

Material and methods:
Methodology: A cross-sectional study was conducted among 323 cancer patients undergoing active treatment at National Teaching Hospital for Oncology, Iraq, selected by systematic sampling. Psychological status was assessed using Hospital Anxiety and Depression Scale (HADS). Clinical and Sociodemographic data were collected through patient interviews and clinical records. Linear regression analysis was employed to determine predictive power of age, cancer type, treatment modality, disease stage, duration, and comorbidities. An XGBoost machine learning classifier was applied to improve prediction of depression and anxiety based on identified clinical and Sociodemographic features.

Results:
Anxiety prevalence was 78.4% (was estimated as severe: 47.7% and moderate: 30.7%), while depression 80.2% (severe: 50.5% and moderate: 29.7%). Age, cancer type, and treatment modality (chemotherapy and radiotherapy) were significant predictors (P<0.001). First year after diagnosis of disease, active treatment period, and disease stage II were identified as critical periods for psychological instability. Patients with chronic comorbidities demonstrated significantly higher psychological levels, indicating a cumulative physical and functional burden. The XGBoost model achieved high predictive performance (accuracy 1.0%, confirming the reliability of Sociodemographic and clinical indicators in predicting high-risk patients.

Conclusions:
Cancer patients are vulnerable to anxiety and depression during early stages of diagnosis and treatment, exacerbated by chronic comorbidities. Incorporating routine psychological screening using HADS, with advanced predictive models like XGBoost, may facilitate early identification of high-risk patients and support targeted interventions to improve quality of life.