Infectious disease prediction model based on optimized deep learning algorithm.

Area of Science:

• Epidemiology and Public Health
• Computational Biology and Bioinformatics
• Data Science and Machine Learning

Background:

• The COVID-19 pandemic highlighted limitations of traditional algorithms in accurately predicting epidemic trajectories due to data complexity.
• Existing models like Autoregressive Integrated Moving Average (ARIMA) capture time-based trends, while Bidirectional Long Short-Term Memory (BiLSTM) networks excel with sequential data.
• Genetic Algorithms (GA) offer optimization for model selection and parameter tuning in complex forecasting tasks.

Purpose of the Study:

• To develop and evaluate a novel hybrid model, GA-BiLSTM-ARIMA, for enhanced prediction of infectious disease outbreaks.
• To assess the predictive performance of the GA-BiLSTM-ARIMA model against standalone BiLSTM and ARIMA models using COVID-19 data.
• To demonstrate the model's capability in improving forecasting accuracy for public health decision-making.

Main Methods:

• Proposed a hybrid forecasting model integrating Genetic Algorithms (GA) for optimization, Bidirectional Long Short-Term Memory (BiLSTM) networks for sequential data analysis, and Autoregressive Integrated Moving Average (ARIMA) for time-series trend capture.
• Utilized COVID-19 case data from Japan for model training and validation.
• Evaluated model performance using standard metrics: Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and R-squared (R²).

Main Results:

• The GA-BiLSTM-ARIMA model achieved superior predictive performance compared to standalone BiLSTM and ARIMA models.
• Specific evaluation metrics for GA-BiLSTM-ARIMA were: RMSE=2,262.42, MAE=1,672.07, MAPE=6.81, and R²=0.9764.
• The hybrid strategy demonstrated robust and higher predictive accuracy in forecasting infectious disease time series.

Conclusions:

• The GA-BiLSTM-ARIMA model effectively combines the strengths of individual algorithms through intelligent optimization, offering a more accurate forecasting tool.
• This advanced hybrid model provides more reliable early warnings and supports the development of effective pandemic prevention and control strategies.
• The findings contribute to a better global response to pandemic challenges by delivering accurate epidemic information to policymakers and the public.