Framed Data Science, Inc.

silicon valley 574

The Organisation

Framed Data, Inc., based in Silicon Valley, San Francisco, is a data science as a service product that helps companies understand their users and improve their business. Rather than focusing on usage analytics, Framed provides predictive machine learning products to identify potential high-value users before they buy and high-risk users before they actually leave.

framed.io

The Challenge

Framed Data have built an automated machine learning platform that takes in user data and predicts when users are going to leave an application. It does this by engineering a feature space out of past user behaviour, and then running a list of selection heuristics to pare down the space. Churn prediction allows a company to determine which of their customers are likely to leave and take steps to prevent this.

Expertise Sought

  • Experience in programming language R and machine learning
  • Experience and ability in feature selection
  • Experience and ability in engineering of high-dimensional datasets
  • Familiarity with statistics and data visualization
  • Familiarity with Git software and GitHub hosting service, or a version control system

The Solution

Philip Spanoudes, MSc Data Science, is working with the company’s CEO and data scientists to improve feature selection and engineering heuristics for the company’s predictive analytics pipeline. They are working together on optimising model accuracy against a subset of sample data. Models that they are creating for this subset will be run against Framed’s production data to evaluate this accuracy. Philip will try to find efficient and effective ways of both enhancing Framed’s churn prediction process and investigating the application of deep learning techniques to their analysis of churn prediction.

Cost

The project was fully funded by Framed Data at a total of £3,000. It is being conducted as part of the MSc Data Science at Lancaster University.

Impact

The research project, as part of a larger project, is helping improve the efficiency of Framed’s churn prediction, delivering an improved approach to feature engineering and selection on high-dimensional datasets at scale as well as improving the feature selection and engineering heuristics for the company’s predictive analytics pipeline. This provides the potential for Framed’s clients to better identify where and why their customers are leaving, and how to prevent them doing so.

Benefits to the company

  • Potential to increase efficiency of Framed’s churn prediction processes
  • Potential to deliver an improved approach to feature engineering and selection on high-dimensional datasets at scale
  • Potential to improve the feature selection and engineering heuristics for the company’s predictive analytics pipeline

Benefits to the university

  • The project is increasing the university’s knowledge of churn prediction systems and how deep learning techniques can be applied to them
  • Provides the University with a new partner for its MSc Data Science

Benefits to society

  • Develop understanding of deep learning techniques
  • Enhance technical skills
  • Hands on experience with Silicon Valley company

Company Feedback

“The calibre of students that we have interviewed at Lancaster has been top notch, and we continue to be impressed not only by their intellect, but also their ability to apply their skills to industry-grade problems, especially in a fast-paced environment like Silicon Valley.”

“We hope to establish a high-quality hiring pipeline of data science candidates from Lancaster University, and Philip is clearly no exception—he has been a great team player on the data team.” Thomson Nguyen, CEO, Framed Data, Inc.

Student Feedback

 “I will be working with Framed Data Inc. investigating the application of deep learning techniques to their analysis of churn prediction. Churn prediction allows a company to determine which of their customers are likely to leave and take steps to prevent this – my project is aimed at finding efficient and effective ways of enhancing this process. I’ll spend most of my time in Lancaster, but will be going to work with Framed Data in San Francisco for a couple of weeks during the summer.

“I’m really excited to have the opportunity to work with cutting-edge data science with Silicon Valley. I’m looking to develop my understanding of deep learning, to deliver real innovation to Framed Data and to enhance both my skills and CV with a great project,”  Philip Spanoudes, MSc Data Science.