Predicting Incidence Time Series Based on LSTM Recurrent Neural Network and Key Feature Analysis Based on Random Forest Algorithm

This paper presents a study on the prediction of the incidence rate time series and the analysis of key features using the LSTM recurrent neural network and the random forest algorithm. Specifically, the study focuses on forecasting the incidence rate for the years 2021-2024 using the ARIMA algorithm and the LSTM recurrent neural network. The ARIMA prediction revealed that the number of diagnosed cases of pulmonary tuberculosis in Liaoning Province reached its lowest point in 2020 over the past decade, showing a 2.4/100,000 decrease in the incidence rate compared to 2019. Furthermore, the LSTM recurrent neural network predicted a continuous decrease in the overall number of cases, indicating a declining trend in the incidence rate. Finally, the random forest algorithm was employed to extract the key features influencing the incidence rate.