The present work aims to analyze the relationship of microbial activity with the physicochemical properties of the soil in banana plantations in Venezuela. Six agricultural fields located in two of the main banana production areas of Venezuela were selected. The experimental sites were differentiated with two levels of productivity (high and low) of the “Gran Nain” banana. Ten variables were selected: total free-living nematodes (FLN), bacteriophages, predators, omnivores, Phytonematodes, saturated hydraulic conductivity, total organic carbon, nitrate (NO3), microbial respiration and the variable other fungi. Subsequently, machine learning algorithms were used. First, the Partial Least Squares-Discriminant Analysis (PLS-DA) was applied to find the soil properties that could distinguish the banana productivity levels. Second, the Debiased Sparse Partial Correlation (DSPC) algorithm was applied to obtain the correlation network of the most important variables. The variable free-living nematode predators had a degree of 3 and a betweenness of 4 in the correlation network, followed by NO3. The network shows positive correlations between FLN predators and microbial respiration (r = 1.00; p = 0.014), and NO3 (r = 1.00; p = 0.032). The selected variables are proposed to characterize the soil productivity in bananas and could be used for the management of soil diseases affecting bananas.