Python Feature Engineering Cookbook

How it works...

In this recipe, we replaced the outliers of one variable of the Boston House Prices dataset from scikit-learn, by the 5^th and 95^th percentiles. We first loaded the data as described in the How it works section of the Trimming outliers from the dataset recipe in this chapter. To replace the outliers, we created a function, that takes the dataframe, the variable name, and the 5^th and 95^th percentiles and uses Numpy's where() to replace the values bigger or smaller than those percentiles by the values of those percentiles.

NumPy's where() scans each observation and if the value is bigger than the 95^th percentile, it replaces it with the 95^th percentile; otherwise, it evaluates whether the value is smaller than the 5^th percentile, ...