Going beyond the textbook: Best practices for creating a DL churn model in healthcare
Who is this presentation for?
- CIOs, CEOs, senior directors, data scientists, and machine learning engineers
It can be a challenge to keep customer churn in check when it comes to providing healthcare insurance. Certain customers can choose to change their service provider at any time; yet the provider isn’t notified until the end of the month when it’s too late for any remedial action. This also makes it difficult to identify the customer’s reason for leaving and limits the ability to collect valuable customer behavior data. For example, it can be impossible to know if a customer is using his health insurance because he’s happy with the service or simply because he has no choice and is sick. On the other hand, if insurance is seldom used, does it mean that she’s unhappy with the service or that she’s just healthy? In each case, it can be difficult to know and take action.
Wovenware’s approach to address customer churn given these limitations was to build a predictive deep learning model to help its insurance client predict which customers are at a higher risk of canceling their subscription during the upcoming month. The goal was to give the business enough time to contact the high-risk clients and address their needs before they canceled their membership.
Leslie De Jesus uses this use case as an example to discuss how churn prediction models are important tools for anticipating customer behaviors and improving retention and based on customer data. You’ll identify three key considerations when creating and maintaining a deep learning churn model for the healthcare industry for it to constantly serve as a tool for preemptive decision making and reduced customer disaffiliation. The three points to consider include healthcare data—claims and impactful features; how to identify the features that impact disaffiliation; data imbalance—how to identify data imbalance and techniques to overcome it; and model validation and statistical bias—techniques to avoid statistical bias.
- General knowledge of deep learning and churn predictive models
What you'll learn
- Discover how deep learning-based churn models provide valuable insights for the healthcare industry and ensure greater customer retention
- Learn to properly address imbalanced data, perform feature engineering, and validate and avoid statistical bias
Leslie De Jesus
Leslie De Jesus is the chief innovation officer at Wovenware. With more than 20 years of expertise in software, product development, and data science, Leslie drives disruptive strategies and solutions, including AI and enterprise cloud solutions, to clients in a variety of markets from healthcare and telco to insurance, education, and defense industries. Leslie is responsible for designing advanced deep learning, machine learning and chatbot solutions, including patented groundbreaking products. One of her biggest strengths is team building, which is the foundation of repetition in the product creation process. Previously, Leslie has held positions such as senior software product architect, CTO, and vice president, product development for key firms.
Leave a Comment or Question
Help us make this conference the best it can be for you. Have questions you'd like this speaker to address? Suggestions for issues that deserve extra attention? Feedback that you'd like to share with the speaker and other attendees?
Join the conversation here (requires login)
Premier Diamond Sponsors
Premier Exhibitor Plus
For conference registration information and customer service
For more information on community discounts and trade opportunities with O’Reilly conferences
For information on exhibiting or sponsoring a conference
For media/analyst press inquires