Data scaling and normalisation explained in 10 minutes with visualizing скачать в хорошем качестве

Data scaling and normalisation explained in 10 minutes with visualizing

In this video, we dive into the world of data scaling techniques, focusing on four popular methods: Standard Scaler, MinMax Scaler, Windsorizing and Robust Scaler. Scaling plays a crucial role in data preprocessing, helping to normalize and standardize features to ensure optimal performance in machine learning models. First, we explore the Standard Scaler, which transforms data to have zero mean and unit variance. We showcase the before-and-after results of applying this technique, highlighting how it can help mitigate the impact of outliers and improve model convergence. Next, we delve into the MinMax Scaler and Windsorizing, which rescale data to a specific range, typically between 0 and 1. By visualizing the data transformations, viewers gain a clear understanding of how this technique maps the original values to the desired range, preserving the data distribution while ensuring consistent scaling. Finally, we introduce the Robust Scaler, a technique designed to handle datasets with outliers. Code from the video https://colab.research.google.com/dri... Github: https://github.com/myredex/ds_tutoria... 00:00 - Introduction 01:11 - Import libraries 02:01 - Generate toy data for scaling 03:08 - How Standard scaler works visualy 05:28 - How MinMax scaler works visualy 06:38 - How Winsorizing data scaling works 08:04 - How Robust scaler works visualy 09:07 - How to work with data scaling during your project 10:37 - Enging