Soil quality assessment is a key component of sustainable land use, agroecosystem monitoring, and the optimization of agrochemical management. The aim of this study is to evaluate the soil quality within the Rozsoshansk community of the Khmelnytskyi region using the Soil Quality Index (SQI) and to develop a statistically sound predictive model based on agrochemical indicators. The analysis was conducted using a set of soil parameters, including pH, Corg, NH4+ NO3-, P2O5, Ca2+ і K+. Spearman correlation analysis was performed, followed by the construction of a multiple linear regression model (R2 = 0.904, Adjusted R2 = 0.900), an assessment of multicollinearity (VIF), and ANOVA. Based on the results of the regression model with seven predictors, together with ANOVA and VIF diagnostics, an optimized model with four predictors (pH, Corg, NO3-, Ca2+) was developed. This refined model demonstrated strong explanatory power (R2 = 0.856, Adjusted R2 = 0.853) and low prediction error. Residual diagnostics indicated deviations from normality and heteroscedasticity; however, robust estimation methods (OLS-HC3 and RLM using HuberT) confirmed the stability of the coefficient estimates. The findings suggest that soil organic carbon, exchangeable calcium, acidity, and nitrate nitrogen are the key indicators of soil quality in the study area. Future work will focus on applying machine learning methods for soil classification, integrating Explainable AI techniques to enhance the interpretability of predictive models.