This research introduces ELM-FISA, an innovative hybrid forecasting model that merges the recently devised parameterless metaheuristic, Fully Informed Search Algorithm (FISA), with an Extreme Learning Machine (ELM). FISA is employed to enhance the weights ( \(w\) ) and biases ( \(b\) ) within the ELM. To assess the forecasting capabilities of this model, two prominent stock indices are employed as the testing domain. The model’s efficacy is gauged using two primary error metrics: ARV and MAPE. The testing involves forecasting the closing prices of each stock index over two distinct time frames: one day ahead and 7 days ahead. For benchmarking, we trained the ELM using another parameter-less metaheuristic, Teaching Learning Based Optimization (TLBO), creating an additional model named ELM-TLBO. Both models undergo a comprehensive evaluation process. The investigational results consistently highlight the superiority of the ELM-FISA model.