Prediction on Predictions with Application to Infectious Disease Spread Analysis


H. Hirose


The 2nd BMIRC International Symposium on Frontiers in Computational Systems Biology and Bioengineering, January 29 - 30, 2013, Fukuoka, Japan.


In observing the gradual increase of patients caused by infectious diseases or the increase of the number of failures of equipment, we anticipate the disease spread widely, and it is crucial to predict the final number of infected patients or failures at earlier stages. To estimate the number of infected patients, the SIR model is commonly used even when the size of observed data is small. Other methods, such as the ordinary differential equation model (ODE), statistical truncated model are also useful to estimate the final number of infected patients. These methods are also applicable to find the increase of the number of failures.
The predicted value for the final number of patients using data until time T becomes a function (trend) of T. We call this L-plot. We, here, consider the use of the L-plot to predict the final number of patients. So far, we have been discussing about the better predictor in the sense that the newly proposed method is superior to other conventional methods. However, in this paper, we try to use all the methods already proposed, and to make a better result than that by using a single method. That is, we will make a prediction using the predicted values already obtained. We call this methodology the PoP, the prediction on predictions.
We here define the decay function using the L-plot. We also propose to use the ensemble method, accepting the majority vote. The PoP in this presentation includes the simple mean value, the decay function, and the ensemble method. By applying the method to the SARS case, we have found that it worked well for early prediction of infectious disease spread.

Key Words



Times Cited in Web of Science:

Cited in Books: