Integrating MANOVA and Decision Tree Classification to Analyze Crop Responses to Fertilizer Nutrients and Soil Conditions

This study examines the relationship between fertilizer nutrients (Nitrogen, Phosphorus, and Potassium), soil conditions, and environmental factors in relation to crop response using secondary data from the Kaggle Repository Crop Recommendation Dataset, comprising 2,200 observations across 22 crop classes. The study was guided by four objectives: to describe the characteristics of fertilizer nutrients and soil conditions, assess differences in nutrient and environmental variables across crop types using Multivariate Analysis of Variance (MANOVA), examine the relationships among environmental variables using Pearson correlation, and evaluate the ability of a Decision Tree model to classify crop types and identify important predictors. The findings showed substantial variability in fertilizer nutrients, particularly Potassium, while soil pH was comparatively stable. MANOVA indicated significant differences in the combined fertilizer nutrient and soil/environmental characteristics across crop types. The correlations among environmental variables were generally weak, suggesting that the variables provide distinct information about crop conditions. The Decision Tree model achieved an overall classification accuracy of 89.24% (Kappa = 0.8873), with Rainfall (19.87%), Potassium (19.20%), and Humidity (17.58%) emerging as the most influential predictors. By integrating MANOVA for assessing multivariate differences with Decision Tree classification for identifying important predictors, the study provides a complementary statistical and machine-learning approach to understanding crop responses. The findings demonstrate the potential of combining these methods to support precision agriculture through evidence-based, crop-specific fertilizer recommendations and improved management of soil and environmental conditions.