When you deal with datasets having more outliers you can convert the normal variate X using the formula
Robust_scaled_feature = (X – median(X))/IQR
We use median and IQR for robust scaling of X
Median and IQR are not affected by outliers. So it is robust
Usage
Use this method when your data contains many outliers, and you don’t want these outliers to influence the scale of your data
Robust scaling uses median and inter quartile range (IQR). This conversion technique is useful when you deal with the features containing more outliers. Use this technique when you want to avoid the heavy influence of outliers in your prediction process.
We use the same data what we used in Z-Score transformation
Load libraries and read data
Separate features(independent variables) and Dependent Variable(churn)
Define Robust Scale Function and convert feature data
Remove the features complains,tariffplan and status
The formula could not be used against the values of these feature we get NaN. So remove and attach the original values
Join the features removed with assign function in another dataframe
Draw boxplot using seaborn and matplot lib
Logistic Regression Model using sklearn lib
Find Coefficients and intercept
Find R-Square
Confusion Matrix
Confusion Matrix Heat Map
Find Accuracy of the model:
Find Precision, recall and F1-score
Classification Report
Regular customers who won’t churn from the telephone company (84.28%)
Default customers point to churning customers from the telephone company (15.72%)
Accuracy works out to 89%
Find ROC-AUC(Area Under Curve)
Calculation of metrics using single program
Comparison of transformation techniques
From the above it is clear that R^2 important metric has improved a lot when we use robust scaling technique for normalizing the feature values
Original data R^2 is 19.94%
z-score R^2 is 20.42%
Robust R^2 is 20.66%