In nanoscience, techniques based on Atomic Force Microscope (AFM) stand as a cornerstone for exploring local electrical, electrochemical and magnetic properties of microelectronic devices at the nanoscale. As AFM's capabilities evolve, so do the challenges of data analysis. With the aim of developing a prediction model for AFM mappings, based on Machine Learning, this work presents a step towards the analysis and benefit of Big Data recorded in the hyperspectral modes: AFM DataCube. The MultiDAT-AFM solution is an advanced 2000-line Python-based tool designed to tackle the complexities of multi-dimensional measurements and analysis. MultiDAT-AFM offers visualization options, from acquired curves to scanned mappings, animated mappings as movies, and a real 3D-cube representation for the hyperspectral DataCube modes. In addition, MultiDAT-AFM incorporates a Machine Learning algorithm to predict mappings of local properties. After evaluating two supervised Machine Learning algorithms (out of the eight tested) for regression, the Random Forest Regressor model emerged as the best performer. With the refinement step, a root mean square error (RMSE) of 0.18, an R2 value of 0.90 and an execution time of a few minutes were determined. Developed for all AFM DataCube modes, the strategy and demonstration of MultiDAT-AFM are outlined in this article for a silicon integrated microelectronic device dedicated to RF applications and analyzed by DataCube Scanning Spreading Resistance (DCUBE-SSRM).