Optimized Long Short-Term Memory Hybrid Model for Infectious Disease Cluster Prediction
This research involves developing a Hybrid prediction model by combining proposed optimized long-sort long-term memory (popLSTM) and K-means (popLSTM-KM) algorithms to predict the number of cases and patterns of the spread of infectious diseases. Other traditional models are unable to predict the number of cases and clusters of spread in detail, which is a serious problem in predicting the spread of infectious diseases. The solution to this problem is to build a hybrid model that can predict the number of cases and clusters simultaneously. So this study proposed a hybrid prediction model that has better accuracy than other hybrid models. This model will be compared with the hybrid models Basic LSTM K-Means and MinMaxSchaler LSTM K-Means. This model tested against four infectious diseases, Monkeypox, Dengue, COVID-19, and Ebola, in four different countries and continents: the United States of America, the Philippines, South Korea, and Guinea. The results showed that the popLSTM-KM model had better cluster accuracy than other hybrid models. The contribution to this study is a decrease in the value of overfitting in the popLSTM-KM model and an increase in the accuracy of the clustering model through modification of the output gate (ot) at the initial data input. The average DBI value of popLSTM-KM is 0.435, this average is better than other hybrid models.