<p> Purpose: Our objective was to analyse the soil characteristics and to develop an appropriate equation for modelling soil moisture content in forest fragments. The study will facilitate predicting soil moisture by simple mathematical equations using readily available soil characteristics. Methods; Study was carried out in Langol reserve forest, Manipur, India during 2022–2024. Stratified random soil sampling was done from two soil depths along the forest fragments. Analysis was done following standard methodologies. 14 soil parameters were used to develop equations for modelling soil moisture. Partial least square regression was analysed in XLSTAT software and Support Vector Regression (SVR) using R software were used for modelling of soil moisture content. Results: Bulk density, moisture content, total organic carbon, soil organic carbon, and soil organic matter values increased with increasing soil depth. 6 soil parameters with high variable importance were selected for modelling. Soil moisture content was found positively correlated to available potassium (R² = 0.486), bulk density (R² = 0.941) and silt (R² = 0.431). The relative significance of each soil variable in the created prediction models is reflected in the variable importance after prediction analysis and found that Bulk density was the most important influencing variable for moisture content in medium and large fragment whereas available potassium in small fragment. The predicted values were close with the observed values with low root mean square errors. Conclusions: The study can help in getting the moisture content of soil samples in inaccessible areas and in data deficient conditions for sustainable soil productivity.</p>