Decision-making requires a combination of intuition, perception and deep thinking skills, which is crucial in personal work and life. Improving decision-making performance is quite beneficial for enhancing people’s competitiveness. This paper predicts the decision-making performance and further analyzes the impact of indoor environment on it. Unlike previous studies, which were mostly empirical, our paper discusses the decision-making performance from a data-driven perspective. We collect data using Chinese chess game and indoor environmental sensors as experimental tools, and propose a hybrid model SA–LSTM, which integrates self-attention (SA) mechanism and long-short term memory (LSTM) network. The results demonstrate that our model can accurately predict the decision-making performance. Compared with other five models, SA–LSTM has lower error and better goodness of fit. Specifically, the root mean square error and mean absolute error of our model are reduced by 11% and 14% compared with the second best performing model LSTM. In addition, SA–LSTM model shows greater robustness. Further interpretability analysis suggests the impact of environmental features on the prediction of decision-making performance, and the SA mechanism enables a deeper exploration of their relationship, thereby emphasizing the significance of environmental variables and enhancing prediction accuracy.