L900 base stations could improve deep coverage and wide coverage

but there were some low effi ciency and unreasonable coverage cells in network. Scene matching was realized automatically for covering urban roads and residential area base station cells

and multi-dimensional modeling were carried out by combining crowdfunding data

DT/CQT data

machine learning grid rate prediction data

SpeedTest data

service data

etc.

so as to select unreasonable coverage scene cells for functional decoupling

so that these cells could be applied to more reasonable coverage scene

improve utilization rate of low effi ciency cells and save equipment investment cost.