<p>Soft soils, known for their low strength and high compressibility, present significant challenges in supporting structural loads due to excessive settlement and limited bearing capacity. Micropiles have been traditionally used to strengthen the ground, but combining them with a soil cement bed (SCB) shows a promising result for even better performance, reducing settlement and increasing load-bearing capacity. Analytical methods for predicting the bearing capacity (<i>Q</i>_<i>u</i>) of micropiles underlying SCB often rely on simplifying assumptions that may not accurately capture the complex interactions between the variables depending on <i>Q</i>_<i>u</i>. To address this limitation, this study conducted laboratory experiments on model micropile groups embedded in a soft soil deposit with an underlying soil cement bed inside a model tank. A comprehensive dataset of 241 data points was collected through load-settlement graphs on 29 experimental tests. These data were used to develop artificial neural network (ANN) models for predicting the bearing capacity of micropile-reinforced soft soil (<i>Q</i>_<i>u</i>). The model incorporated five key dimensionless parameters affecting bearing capacity: length (<i>L</i>) of micropile, number (<i>n</i>) of micropiles, spacing (<i>s</i>) between the micropiles, thickness (<i>H</i>) of soil cement bed and settlement (<i>s</i>_<i>t</i>) of soft soil. The ANN-predicted results were validated against experimental outcomes, showcasing the effectiveness and reliability of the ANN approach in predicting the load-bearing behavior of micropile-SCB-soft soil systems. The dataset generated and analysed during this study is fully available within the manuscript, ensuring transparency and reproducibility for future research.</p>