Fig. 2From: Exploring supervised machine learning approaches to predicting Veterans Health Administration chiropractic service utilizationResults of principal component analysis of 158 feature inputs. Separability of four classes representing one-year chiropractic service utilization was poor when projected into two-dimensional space based on first two principal components. The explained variance ratio for the target variable as a function of the total number of principal components shows limited evidence of a strong influence on the variance between individual principal components as a predictor of the labelBack to article page