<p>Tunnel engineering faces significant challenges due to the complexity and variability of geological conditions. In such contexts, the timely and appropriate adjustment of shield construction parameters (SCP) to match current geological conditions is crucial for ensuring safe and efficient construction. This paper proposes a self-adaptive intelligent decision-making method for risk control in shield tunnel construction, which integrates a multilayer perceptron (MLP) with a multi-objective optimization (MOO) algorithm. Specifically, an MLP combined with particle swarm optimization (PSO) is employed to predict the optimal SCP. These predicted parameters, along with stratum mechanical properties and tunnel depth, serve as inputs for forecasting two critical performance indicators: maximum surface settlement (<i>MSS</i>) and driving speed (<i>DS</i>). The predictive model, coupled with the MOO algorithm, is then utilized to enable dynamic feedback control of the SCP during construction. To demonstrate the applicability of the proposed method, a case study of Qingdao Metro Line 6 is presented. The results indicated that based on the proposed PSO-MLP-non-dominated sorting genetic algorithm II (NSGA-II), both <i>MSS</i> and <i>DS</i> are effectively improved. The adaptive decision-making between SCP and geological conditions can be realized.</p>