Background: Colorectal cancer (CRC) remains one of the leading causes of cancer-related mortality worldwide, with growing evidence linking risk to lifestyle and dietary factors. However, nutrition-related exposures have rarely been integrated into existing CRC risk prediction models. This study aimed to develop and validate a lifestyle-based 10-year CRC risk prediction model using longitudinal data from a large-scale population-based screening cohort to facilitate early risk stratification and personalized screening strategies. Methods: Data were obtained from 21,358 individuals participating in a CRC screening program in Shanghai, China, with over 10 years of active follow-up until 30 June 2021. Of these participants, 16,782 aged ≥40 years were used for model development, and 4576 for external validation. Predictors were selected using random survival forest (RSF) and elastic net methods, and the final model was developed using Cox regression. Machine learning approaches (RSF and XGBoost) were additionally applied for performance comparison. Model performance was evaluated through discrimination, calibration, and decision curve analysis (DCA). Results: The final model incorporated twelve predictors: age, gender, family history of CRC, diabetes, fecal immunochemical test (FIT) results, and seven lifestyle-related factors (smoking, alcohol use, body shape, red meat intake, fried food intake, pickled food intake, and fruit and vegetable intake). Compared to the baseline demographic-only model (C-index = 0.622; 95% CI: 0.589–0.657), the addition of FIT improved discrimination, and further inclusion of dietary and lifestyle variables significantly enhanced the model’s predictive accuracy (C-index = 0.718; 95% CI: 0.682–0.762; ΔC-index = 0.096, p = 0.003). Conclusions: Incorporating dietary and lifestyle variables improved CRC risk stratification. These findings highlight the value of dietary factors in informing personalized screening decisions and providing an evidence-based foundation for targeted preventive interventions.