Doing Predictive Analytics in Practice

Mark your calendar! Our speaker this time is Karim Maarouf, senior data scientist at Teradata. Karim will share with us some of his experience developing large scale predictive analytics (machine learning applied to predicting future events to optimize for some business value) for some of the world's largest telecommunication providers. Whether you are a practicing data scientist looking to gain new insights or a new comer to the field looking to get a feeling of how predictive analytics is done in practice, this talk is for you.

Biography:

Karim Maarouf is a senior data scientist at Teradata Egypt. He has over 5 years of experience helping leading companies in Egypt and the region leverage analytics to enhance different aspects of their business. Karim’s work is focused on predictive marketing. He has worked with different organizations on building analytical roadmaps and implementing analytical solutions that have an impact on their P&L. Throughout his work, Karim also handled several demand creation activities and enjoys speaking about the latest trends in technology and data science whether related to Teradata or otherwise.

Karim started out as a researcher in computer science and engineering and holds a master’s degree from the German University in Cairo. He authored several papers in the field of video coding.

Talk outline:

Practical Predictive Analytics: Organizations utilize predictive analytics, including a variety of statistical and machine learning techniques, to study historical data and make predictions relevant for their business. Common examples include predicting customer churn (customers canceling their subscription), finding customers who are likely to buy new products and finding instances of fraud. There are many approaches and considerations that need to be taken into account in order to successfully apply predictive analytics at an organization. This talk aims to bridge the gap between theory and practice when it comes to applying predictive analytics. We will discuss:

1. Some approaches to performing predictive modeling to provide business value.

2. We will also discuss how to avoid common pitfalls and some tips and tricks to simplify building a good predictive model.

3. Finally, we will discuss how to evaluate the accuracy and usability of a predictive model.

To keep things interesting and to guide us through the talk, we will examine all of these details from the point of view of a common application of predictive modeling, which is predicting churn for a telecommunications provider.