Study on the prediction of bloodstain age based on near infrared spectroscopy and long short-term memory algorithm

How to predict the age of bloodstains rapidly at the crime scene is of great significance for the investigation of violent crimes. In this paper, the near infrared (NIR) spectral data of human bloodstain samples on different substrates were collected by using a hand-held NIR spectrometer. First derivative pre-processing and a long short-term memory algorithm (LSTM) were used to establish the prediction model of bloodstain age. The experimental results showed that the determination coefficient, R 2 , of the test set using the LSTM algorithm was 0.95, which was much higher than that of the partial least squares (PLS) algorithm 0.71 and random forest (RF) algorithm 0.84. Meanwhile, the root mean square error of the LSTM algorithm was 2.13 days, which was lower than that of the PLS algorithm 5.24 days and RF algorithm 3.85 days. The above results indicated that NIR spectroscopy combined with the LSTM algorithm could predict the bloodstain age efficiently, non-destructively, and accurately. It provided a new technical reference for the rapid prediction of bloodstain age in criminal cases.