| Summary: | Engagement in physical activity (PA) is decreasing due to the lack of time, facility, and knowledge - three commonly cited barriers to PA. Novel COVID -19 has further disrupted engagement in PA for many by limiting access to facilities and qualified exercise professionals. Virtual home-based bodyweight fitness software may serve as a good intervention in resolving these barriers. However, commercial fitness apps rarely feature individualized exercise programming, nor has past literature shown how to develop one. PURPOSE: To demonstrate a novel algorithm and develop a nonprofit open-source exercise program generating app. METHODS: Using ACSM recommendations, an algorithm was designed that stresses basic exercise principles. Figure 1 shows the recommender system of six processes in our design. The first process is the pre-PA screening. From second to fourth processes, information of goal, time, and exercise level is collected generating up to twenty-seven associated exercise patterns. Exercises are then randomly deployed into these patterns to maximize variability. Lastly, a detailed report is generated. Based on this design flowchart and a detailed algorithm, Node.js, Express framework, Mongo database, and Azure cloud were used to program the app. RESULTS: Over 200 exercises and their instructions indexed in the database were randomly deployed into these patterns each time when a program is generated with over approximate 2 * 1028 variable outputs. The final app was deployed at cloud: home-exercise.azurewebsites.net. Detailed algorithm and source code were publicly hosted here: github.com/columbia-dsi/fit. CONCLUSION: An individualized home-based bodyweight exercise program generating app was created, which offers a potential solution to overcome major public PA barriers. Open source code is also available as the reference for future researchers to develop a more intelligent system and investigate how virtual training tools can help increase PA.
|
|---|