<p>Accurate prediction of foundation pit settlement (SoFP) remains a challenge due to limitations in real-time excavation depth tracking. This study introduces a new hybrid method that boosts prediction accuracy by combining support installation records with metaheuristic optimization. A key goal is to accurately record the installation of supports. By analyzing the gaps between supports in terms of depth and time, a more precise estimate of the excavation depth rate can be made. The study compares three methods for forecasting SoFP, with a focus on merging the Fire Hawk and Prairie Dog hybrid optimization techniques with the Adaptive Neuro-Fuzzy Inference System (ANFIS) model. It has been shown that using a hybrid approach can increase the accuracy and reliability of SoFP estimates, ultimately improving the safety and efficiency of excavation operations. The study considers various factors that impact settlement, including the internal friction angle (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\varphi\)</EquationSource> <EquationSource Format="MATHML"><math> <mi>φ</mi> </math></EquationSource> </InlineEquation>), cohesion (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(c\)</EquationSource> <EquationSource Format="MATHML"><math> <mi>c</mi> </math></EquationSource> </InlineEquation>), bulk density (<InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(k\)</EquationSource> <EquationSource Format="MATHML"><math> <mi>k</mi> </math></EquationSource> </InlineEquation>), Poisson’s ratio (<InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(\nu\)</EquationSource> <EquationSource Format="MATHML"><math> <mi>ν</mi> </math></EquationSource> </InlineEquation>), void ratio (<InlineEquation ID="IEq5"> <EquationSource Format="TEX">\(e\)</EquationSource> <EquationSource Format="MATHML"><math> <mi>e</mi> </math></EquationSource> </InlineEquation>), variations in the water table, permeability coefficient (<InlineEquation ID="IEq6"> <EquationSource Format="TEX">\(K\)</EquationSource> <EquationSource Format="MATHML"><math> <mi>K</mi> </math></EquationSource> </InlineEquation>), number of supports, and excavation depth. Based on the uncertainty at the 95% confidence level (U₉₅), the ANFIS(PDA) yielded U₉₅ index values of 5.2814 and 5.2814 for the assessment and learning stages, accordingly. The U₉₅ values for the ANFIS(FHA) were 6.5733 and 5.4385, respectively. These values remained consistent throughout the project duration. After considering assessment variables, logical inference, and rating level, it has been determined that all models are deemed reliable and dependable. However, the ANFIS(PDA) model has a slight edge over the other model in terms of its objective, due to its more effective exploration of feature space and faster convergence during optimization, which results in more accurate parameter tuning.</p>