<p>This study proposes an improved hybrid soft-computing paradigm for estimating the optimum moisture content (OMC) and maximum dry density (MDD) of coarse and fine-grained soils. The proposed method integrates the adaptive opposition slime mould algorithm (AOSMA) and extreme learning machine (ELM) to develop an efficient prediction model, ELM-AOSMA. The AOSMA is an improved version of the standard slime mould algorithm (SMA) that enhances predictive accuracy by maximising the exploitation capability of the search process. The proposed model was trained and validated using a database compiled from the literature and newly conducted laboratory experiments. The performance of the ELM-AOSMA model was compared with five additional hybrid ELMs constructed using different swarm intelligence algorithms, along with the standard SMA. Based on the results, the developed ELM-AOSMA model attained the most accurate predictions of compaction parameters with 97.78% accuracy (as per the R² value) in the testing phase. Moreover, parametric analysis was performed to demonstrate the influence of particle sizes and compaction energy on the compaction characteristics of soils. Overall, the ELM-AOSMA model can serve as an efficient tool for rapid estimation of OMC and MDD. A graphical user interface was also developed and attached as supplementary material.</p>