In order to improve the prediction accuracy of the rabbit house environment parameters, solve the coupling relationship between environmental parameters ignored in traditional predict method, and reduce the cost of rabbit house environmental control, a multivariable environmental prediction sequence to sequence model of rabbit house based on Long Short-Term Memory was proposed. Double-layer LSTM was used as the encoder and decoder of the Seq2Seq structure to improve the characterization ability and prediction accuracy of the environmental parameter prediction model. The Seq2Seq structure can not only effectively extract the time correlation of the rabbit house environmental parameter sequence itself, but also can mine the coupling relationship between the parameters. The model was used to test and predict the data of temperature, humidity and carbon dioxide concentration in the rabbit house which in a rabbit farm in Shengzhou City, Zhejiang Province. The results showed that the multi-parameter prediction model of the rabbit house environment achieved good prediction performance. Compared with standard LSTM model and standard SVM model, the prediction accuracy of temperature is improved by 28.41% and 48.60%, the prediction accuracy of humidity is improved by 9.84% and 56.08%, and the prediction accuracy of carbon dioxide concentration is improved by 5.39% and 11.19%. The experimental results showed that the proposed multivariable environmental prediction model of rabbit house not only had good forecasting effect, but also can meet the needs of accurate of prediction of rabbit house environmental data.