|
Abstract:
|
Postmortem Interval (PMI) is the time that has elapsed since a person or animal has died. In forensic sciences, if the time is unknown, a number of medical/scientific techniques need to be used to determine it. Recently, statistical models have been applied in predicting postmortem interval based on the dynamic change in microbial community of the corpses. However, there lacks of accuracy in predicting PMI due to various types of noises. In this research, we propose to employ a data mining method, regularized random forest model with leave one out sampling technology, to predict PMI. Concurrently, the external information about the corpses is also considered in the predication procedure. We demonstrate that the new approach excels all available models in terms of prediction accuracy via published mouse data.
|
