diff --git a/Feature Engineering- Missing Values Day 2.ipynb b/Feature Engineering- Missing Values Day 2.ipynb index c7cc2fb..aaf0887 100644 --- a/Feature Engineering- Missing Values Day 2.ipynb +++ b/Feature Engineering- Missing Values Day 2.ipynb @@ -1215,6 +1215,23 @@ "sns.boxplot('Age_end_distribution',data=df)" ] }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Advantages\n", + "1. Easy to implement\n", + "2. Captures the importance of missingess if there is one\n", + "\n", + "### Dsiadvantages\n", + "1. Distorts the original distribution of the variale\n", + "2. If missingess is not importrant, it may mask the predictive power of the original variable by distorting its distribution\n", + "3. if the number of NA is big, it will mask the true outliers in the distribution\n", + "4. if the number of NA is small, the replaced NA may be considered an outlier and pre-processed in a subsequent feature engineering\n", + "\n", + " " + ] + }, { "cell_type": "code", "execution_count": null,