How to transform a dataset for a machine learning model
In this course, we are going to focus on pre-processing techniques for machine learning.
What you’ll learn
- How to fill the missings in numerical and categorical variables.
- How to encode the categorical variables.
- How to transform the numerical variables.
- How to scale the numerical variables.
- Principal Component Analysis and how to use it.
- How to apply oversampling using SMOTE.
- How to use several useful objects in scikit-learn library.
Course Content
- Introduction –> 5 lectures • 15min.
- Data cleaning –> 7 lectures • 57min.
- Encoding of the categorical features –> 5 lectures • 46min.
- Transformations of the numerical features –> 7 lectures • 43min.
- Pipelines –> 3 lectures • 32min.
- Scaling –> 3 lectures • 22min.
- Principal Component Analysis –> 3 lectures • 20min.
- Filter-based feature selection –> 9 lectures • 1hr.
- A complete pipeline –> 1 lecture • 19min.
- Oversampling –> 3 lectures • 21min.
- General guidelines –> 1 lecture • 1min.
Requirements
In this course, we are going to focus on pre-processing techniques for machine learning.
Pre-processing is the set of manipulations that transform a raw dataset to make it used by a machine learning model. It is necessary for making our data suitable for some machine learning models, to reduce the dimensionality, to better identify the relevant data, and to increase model performance. It’s the most important part of a machine learning pipeline and it’s strongly able to affect the success of a project. In fact, if we don’t feed a machine learning model with the correctly shaped data, it won’t work at all.
Sometimes, aspiring Data Scientists start studying neural networks and other complex models and forget to study how to manipulate a dataset in order to make it used by their algorithms. So, they fail in creating good models and only at the end they realize that good pre-processing would make them save a lot of time and increase the performance of their algorithms. So, handling pre-processing techniques is a very important skill. That’s why I have created an entire course that focuses only on data pre-processing.
With this course, you are going to learn:
- Data cleaning
- Encoding of the categorical variables
- Transformation of the numerical features
- Scikit-learn Pipeline and ColumnTransformer objects
- Scaling of the numerical features
- Principal Component Analysis
- Filter-based feature selection
- Oversampling using SMOTE
All the examples will be given using Python programming language and its powerful scikit-learn library. The environment that will be used is Jupyter, which is a standard in the data science industry. All the sections of this course end with some practical exercises and the Jupyter notebooks are all downloadable.