In the 3G era, the optimal parameter combination can be selected from 13 radio parameter combinations based on experience. In the 4G and 5G eras, it is almost impossible to find out the optimal value of pattern by relying on past experience in the face of hundreds of or even thousands of parameter combinations. According to the feedback of SoftBank, they once tried to manually adjust a pattern, but the operation efficiency was low and the risk was high. Adjustment of a single pattern required about one week long and this way of adjustment is impractical.
In this solution, the random forest algorithm is used for modeling and prediction. The live network data is collected, analyzed, sorted, and marked. Then the data is sent to the AI reasoning platform to calculate the optimal initial value by using the random forest model. Based on the optimal initial value, the optimal parameter combination is obtained in a short term by means of iterative optimization.
This solution is applied on the live network of operator S in country J. It helps find the optimal initial value within one day, and complete iterative optimization and optimal combination of configuration parameters within one week. This provides an increase of 17% in network traffic and an increase of 18.4% in user capacity.