Predictive Analytics using Python

This course provides you with the skills to build a predictive model from the ground up, using Python.

You will learn the full lifecycle of building the model. First, you'll understand the data discovery process and discover how to make connections between the predicting and predicted variables. You will also learn about key data transformation and preparation issues, which form the backdrop to an introduction in Python for data analytics.

Through the analysis of real-life data, you will also develop an approach to implement simple linear and logistic regression models. These real-life examples include assessments on customer credit card behavior and case studies on sales volume forecasting.

This course is the first in the MicroMasters program and will prepare you for modeling classification and regression problems with statistical and machine learning methods.

A predictive exercise is not finished when a model is built. This course will equip you with essential skills for understanding performance evaluation metrics, using Python, to determine whether a model is performing adequately.

Specifically, you will learn:

• Appropriate measures that are used to evaluate predictive models
• Procedures that are used to ensure that models do not cheat through, for example, overfitting or predicting incorrect distributions
• The ways that different model evaluation criteria illustrate how one model excels over another and how to identify when to use certain criteria

This is the foundation of optimising successful predictive models. The concepts will be brought together in a comprehensive case study that deals with customer churn. You will be tasked with selecting suitable variables to predict whether a customer will leave a telecommunications provider by looking into their behaviour, creating various models, and benchmarking them by using the appropriate evaluation criteria.

This course is running for the final time – if you wish to sign up then you must do so by 5 March 2022.

In this course, you will learn three predictive modelling techniques - linear and logistic regression, and naive Bayes - and their applications in real-world scenarios.

The first half of the course focuses on linear regression. This technique allows you to model a continuous outcome variable using both continuous and categorical predictors. This technique enables you to predict product sales based on several customer variables.

In the second half of the course, you will learn about logistic regression, which is the counterpart of linear regression, when the response variable is categorical. You will also be introduced to naive Bayes; a very intuitive, probabilistic modeling technique.

This course is running for the final time – if you wish to sign up then you must do so by 15 April 2022.

This course will give you an overview of machine learning-based approaches for predictive modelling, including tree-based techniques, support vector machines, and neural networks using Python. These models form the basis of cutting-edge analytics tools that are used for image classification, text and sentiment analysis, and more.

The course contains two case studies: forecasting customer behaviour after a marketing campaign, and flight delay and cancellation predictions.

You will also learn:

• Sampling techniques such as bagging and boosting, which improve robustness and overall predictive power, as well as random forests
• Support vector machines by introducing you to the concept of optimising the separation between classes, before diving into support vector regression
• Neural networks; their topology, the concepts of weights, biases, and kernels, and optimisation techniques