<p>Actual evapotranspiration (ETa) is a vital part in hydrologic cycle, and the analysis of influencing factors and prediction of ETa are of great relevance for the efficient utilization of water resources and sustainable development of agriculture in the irrigated areas. Given that previous research methods have issues with poor explanatory power or low predictive accuracy, an improved random forest model was applied to ascertain the main influencing factors of ETa, and a prediction model for ETa with meteorological and human factors based on attention mechanism, BP neural network and LSTM neural network was established in irrigated areas in this paper. The results showed that temperature and solar radiation were the main influencing factors of ETa, and the influence of human factors could not be ignored. Compared with Attention-LSTM model and BP model based on meteorological factors, the performance of Attention-LSTM-BP model increased by 31.8 and 35.8% and 75.2 and 78.0% in MAE and RMSE, respectively. Moreover, after integrating human factors, the performance of Attention-LSTM-BP model was further improved, with R² increase of 9.7%.</p>