|
| Abstract: |
| A new health concern in recent periods has seen the evolution of uncertain sedentary behavior. Remaining sedentary for extended durations is regarded as a notable hazard across various adult age brackets, especially the excessive dependence on automobiles for transportation. Throughout the active period, monitoring seating habits has been made easier by sensors. Nevertheless, there exists a disagreement among professionals regarding the most suitable quantifiable criteria for encompassing the comprehensive data on sedentary behavior throughout the day. Owing to variations in measurement methodologies, data analysis approaches, and the lack of essential outcome indicators such as the total sedentary duration, the assessment of sedentary patterns in numerous research investigations was considered unfeasible. The research suggested fleeting granularity distinguish occurrences of regular human activities. Sophisticated units (essential cells) acquire multivariate transitory information. Frequent Behavior Patterns (FBPs) can be identified with a estimation of timeframe using our proposed scalable algorithms that employ collected widespread multivariate data (fleeting granularity). The research outcome, supported by rigorous analyses on two validated datasets, mark a significant progression. In the final stages of the study, a stacked Long Short-Term Memory (LSTM) model was utilized to replicate and forecast repetitive sedentary behavior patterns, leveraging data from the preceding six-hour window blocks of sedentary activity. The model effectively replicated state traits, previous action sequences, and duration, attaining an impressive 99% accuracy level as assessed through RMSE, MSE, MAPE, and r-correlation metrics. |
| Key words: health care sedentary behavior LSTM |
| DOI：10.11916/j.issn.1005-9113.2023052 |
| Clc Number:TP391 |
| Fund: |
