Integrated learning method for predicting lethal effect of short-term exposure of nerve poison

The invention discloses an integrated learning method for predicting the lethal effect of short-term exposure of nerve poison. According to the invention, a heterologous data set covering different test animals and a plurality of exposure pathways is established, and a plurality of toxicity mechanisms are involved. According to the method, the research normal form that only molecular structure characteristics are considered and biological exposure characteristics are ignored in traditional machine learning modeling is broken through, different test animals and multiple exposure ways are comprehensively considered and are subjected to one-hot coding, the molecular structure characteristics are coupled, and an integrated learning prediction model based on hard voting combination is developed. The neural poison short-term exposure lethal effect prediction model established by the invention has relatively high internal robustness and external prediction ability, has a clear application domain, is simple and conve