ADS Capstone Chronicles Revised

18

Figure 11 Heatmap of Correlation Coefficients for Numeric and Ordinal Features

The Pandas .corr() function is not compatible with the binary features presented in the data. Rather, the binary features will be correlated versus the Glucose Value to determine if there is any relation with a point biserial correlation calculation. However, Glucose Value is not considered the target variable as this data science project has unsupervised learning methods to create a recommendation system. Table 2 presents the point biserial correlation of each binary feature related to Glucose Values, a continuous numeric feature. These calculations do not present any strong relationships; negative or positive. These results are expected as these binary features are unique to each patient and represent a variety of health data scenarios.

220