Researchers from Worcester Polytechnic Institute (WPI) from Worcester, Massachusetts, and the University of Connecticut have developed a smartphone app that uses machine learning to predict eating patterns to provide potential interventions for users hoping to lose weight. Findings were presented at the annual symposium for the American Medical Informatics Association.

Although the number of weight-loss apps currently available nears 30,000, these apps are focused on tracking activity, calories and weight. In the SlipBuddy app, factors of stress and eating are taken into account to build a personalized intervention system using behavioral strategies for improved outcomes. The app integrates these factors and machine learning to identify which factors cause the user to overeat.

"I'm very hopeful that what we're doing will make a big difference," said Bengisu Tulu, associate professor in WPI's Foisie Business School. "Most weightloss apps are all about tracking something—tracking your calories, tracking your blood glucose, tracking your steps. This goes beyond that. We're using machine learning to make this about intervention."

In the study, researchers enrolled 16 adults who were overweight and supplied them with the app for a month. After the study period, nine participants had lost an average of five pounds, three remained the same and four gained an average of two pounds.

"Mobile technology, which is ubiquitous today, has the capacity to deliver evidence-based weight loss interventions with lower cost and user burden than traditional intervention models," said Carolina Ruiz, associate professor of computer science at WPI.

“This is truly an interdisciplinary project that pushes the boundaries in obesity research," said Ruiz. "The use of machine learning algorithms to uncover accurate predictive patterns of behavior allows our app to deliver user-centric, evidence-based, personalized approaches to prevent overeating, which will have a positive impact on combating obesity."