The Moments in Data Science Part 2 Application of Skewness and Kurtosis

Skewness , kurtosis applications in Data science.

Skewness:
Skewness is like the horizontal pull on your data. It tells you how spread out the data is along the x-axis.
Imagine a histogram or a distribution plot. When it’s perfectly symmetric (like a well-behaved bell curve), the skewness is zero. The tails on both sides are equal in length.
But when there’s asymmetry, giving the distribution a leaned, squished-to-one-side look, we say it’s skewed.
There are three types of skewness:
Positive skewness: The tail stretches out to the right (longer right tail).
Negative skewness: The tail stretches out to the left (longer left tail).
Zero skewness: A symmetric distribution with values evenly centered around the mean.

Kurtosis:
Kurtosis is all about the vertical pull or the peak’s height. It tells you how tall and pointy the central peak of your distribution is.
A normal distribution (the beloved bell curve) has a kurtosis of 3. When kurtosis is greater than 3, the peak is sharper (more pointy), and when it’s less than 3, the peak is flatter.

Types of kurtosis:
Leptokurtic: Kurtosis more than 3 (sharper peak).
Mesokurtic: Kurtosis = 3 (normal distribution).
Platykurtic: Kurtosis less than 3 (flatter peak).
Practical Applications:
#data Transformation:
Highly skewed data might benefit from normalization or scaling techniques to make them resemble a normal distribution. This aids in model performance.
For example, in finance, stock returns are often skewed. Transforming them can improve model accuracy.
Outlier Detection:
Skewness and kurtosis help identify outliers. Extreme values can distort #statisticalmodeling analyses.
Detecting outliers is crucial for robust modeling.
Model Selection:
Some machine learning algorithms (like linear models) assume normality.
Understanding skewness and kurtosis guides your choice of algorithms.