Hybrid machine-learning algorithms for predicting the number of road accident patients in Golestan Province using time-series data

Background: This study employs hybrid machine learning algorithms to analyze the number of patients resulting from road accidents in hospitals in Golestan Province over a five-year period from March ۲۰۲۰ to March ۲۰۲۵. It also aims to predict the number of traffic accident patients from March ۲۰۲۵ to March ۲۰۲۷. Methods: In this retrospective study, a five-year dataset covering March ۲۰۲۰ to March ۲۰۲۵ was used. Hybrid machine-learning algorithms, including SARIMA-LSTM and CNN-LSTM, were applied for analysis and prediction in comparison with the traditional Seasonal Autoregressive Integrated Moving Average (SARIMA) model. The performance of these algorithms was evaluated using root mean square error (RMSE), mean absolute error (MAE), and root mean square logarithmic error (RMSLE). Results: The observed number of patients referred following road accidents in Golestan Province increased from ۱۳,۶۷۹ in ۲۰۲۰ to ۲۰,۳۲۳ in ۲۰۲۵. A statistically significant difference was observed in patient numbers between men and women (P-Value < ۰.۰۰۱), with five-year increases of ۴۵% for men and ۶۰% for women. For predictions covering March ۲۰۲۵-March ۲۰۲۷, the CNN-LSTM model achieved the lowest error metrics (MAE (%) = ۴.۵, RMSE = ۶.۵۰, RMSLE = ۰.۲۱), followed by SARIMA-LSTM (MAE (%) = ۸.۶۰, RMSE = ۱۲.۴۵, RMSLE ۰.۶۵= ۰.۱۵), whereas the conventional SARIMA model exhibited the highest errors (MAE (%) = ۱۲.۶۹, RMSE = ۱۵.۳۷, RMSLE = ۰.۷۸). Conclusion: These findings may assist policymakers and health managers in improving health services through optimal resource allocation and enhanced planning. Furthermore, the use of machine-learning models in hospitals is recommended to support the management and prediction of traffic-accident-related patient volumes.

کلیدواژه ها:

Machine Learning ، Time Series ، Road Accident Patients ، Predicting