Operationalizing Machine Learning at Comcast

Technology Category

Analytics & Modeling - Machine Learning
Analytics & Modeling - Predictive Analytics

Applicable Industries

Telecommunications

Applicable Functions

Discrete Manufacturing
Quality Assurance

Use Cases

Predictive Maintenance
Real-Time Location System (RTLS)

Services

Data Science Services
System Integration

The Challenge

Comcast, a multinational mass media company, was facing challenges in building models on complete production datasets to improve the accuracy of their predictive analytics. They needed a solution to run models on complete production datasets as opposed to relying on sampling. The company was also looking for ways to operationalize and scale data science to match the company's volume of operations. They were dealing with issues such as extracting and integrating real-time production data that comes in different formats, using heavy computation to transform raw data into usable data sources, providing timely responses to a great number of prediction requests and continuously updating models with the latest data to keep predictions accurate.

About The Customer

Comcast is a multinational mass media company with multiple lines of business. Their operations span across high-speed Internet, feature and TV film productions, cable channels, IP telephony, and home security and automation services. The company has a large-scale system that generates massive amounts of data from tens of millions of customers and hundreds of millions of networked devices. They use this data to deliver personalized content to diverse TV audiences across the US, improve customer care, build resiliency into its products and reduce truck rolls for the service technicians.

The Solution

Comcast adopted H2O.ai Machine Learning and Datameer to overcome their challenges. They built predictive models to prevent avoidable truck rolls, which are appointments at a customer's home or business that could be avoided by using simple fixes. They used Datameer for data engineering and H2O to split the data into training and test datasets. Once the training was completed, the data was validated using the testing dataset. They also used gradient-boosted decision trees to predict the popularity of a particular TV show or film 24 hours in advance and make recommendations to the viewers. They used clustering algorithm to assemble the data to form customer experience groups. They also used data analysis to make more resilient and reliable products.

Operational Impact

Cost savings and efficiency gains in reducing avoidable truck rolls.
Accurate predictions of trending content to offer personalized suggestions and browsing options to TV audiences.
Improved metrics to evaluate customer experience.