Scholarlink

Research hypothesis: Time series modeling can provide accurate daily outpatient visits forecasting of the clinic. A combined approach of using two forecasting models: Artificial Neural Networks and the Seasonal Autoregressive Integrated Moving Average (SARIMA) stochastic model to approximate nonlinear and linear relationships, respectively, increases forecasting accuracy by taking into account the specifics of outpatient visits on weekdays and weekends, or the so-called the day week effect. Method: Modeling the forecast of stochastic time series. Dataset: The original single time series consists of 787 cases—the number of daily outpatient visits (persons) to the Kazan clinic (Russia) in the time periods: 01.01.2022–02.27.2024. Results: The validity of the hypothesis is confirmed. By combining the complementary capabilities of Artificial Neural Networks and the SARIMA model, there was an improvement in the accuracy of forecasting daily outpatient visits in the short term. The weekend forecast results demonstrated that Mean Absolute Percentage Error is significantly reduced when using neural networks. The obtained Mean Absolute Percentage Error estimates for working days of the week showed the best results when applying the SARIMA model. Conclusion: Artificial Neural Networks time series forecasting models in combination with the SARIMA model allow better to specify of the time series and more accurate modeling of the short-term forecast.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Clinic Daily Outpatient Visits Forecasting Using a Combination Method Based on Neural Networks and SARIMA Model

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Kirshin Igor,
  • Kirshin Dmitry,
  • Minnullin Ramil

摘要

Research hypothesis: Time series modeling can provide accurate daily outpatient visits forecasting of the clinic. A combined approach of using two forecasting models: Artificial Neural Networks and the Seasonal Autoregressive Integrated Moving Average (SARIMA) stochastic model to approximate nonlinear and linear relationships, respectively, increases forecasting accuracy by taking into account the specifics of outpatient visits on weekdays and weekends, or the so-called the day week effect. Method: Modeling the forecast of stochastic time series. Dataset: The original single time series consists of 787 cases—the number of daily outpatient visits (persons) to the Kazan clinic (Russia) in the time periods: 01.01.2022–02.27.2024. Results: The validity of the hypothesis is confirmed. By combining the complementary capabilities of Artificial Neural Networks and the SARIMA model, there was an improvement in the accuracy of forecasting daily outpatient visits in the short term. The weekend forecast results demonstrated that Mean Absolute Percentage Error is significantly reduced when using neural networks. The obtained Mean Absolute Percentage Error estimates for working days of the week showed the best results when applying the SARIMA model. Conclusion: Artificial Neural Networks time series forecasting models in combination with the SARIMA model allow better to specify of the time series and more accurate modeling of the short-term forecast.