Coverage indicator prediction method, model training method and apparatus, device and medium
Provided is a coverage indicator prediction method. The method includes: obtaining a wireless cell feature of a wireless cell to be predicted, a geographical environment feature of the wireless cell to be predicted, a grid geographical environment feature, and a feature of a wireless propagation path from the wireless cell to be predicted to a corresponding grid, where grids are obtained by dividing a designated region; and predicting, according to the wireless cell feature of the wireless cell to be predicted, the geographical environment feature of the wireless cell to be predicted, the grid geographical environment feature, and the feature of the wireless propagation path from the wireless cell to be predicted to the corresponding grid, a coverage indicator value of the grids using a trained coverage indicator prediction model. Coverage indicator prediction apparatus, model training method and apparatus, electronic device, and computer-readable storage medium are also provided.
1. A coverage indicator prediction method, comprising:
obtaining a wireless cell feature of a wireless cell to be predicted, a geographical environment feature of the wireless cell to be predicted, a grid geographical environment feature, and a feature of a wireless propagation path from the wireless cell to be predicted to a corresponding grid, wherein grids are obtained by dividing a designated region; and
predicting, according to the wireless cell feature of the wireless cell to be predicted, the geographical environment feature of the wireless cell to be predicted, the grid geographical environment feature, and the feature of the wireless propagation path from the wireless cell to be predicted to the corresponding grid, a coverage indicator value of the grids using a trained coverage indicator prediction model,
wherein before predicting, according to the wireless cell feature of the wireless cell to be predicted, the geographical environment feature of the wireless cell to be predicted, the grid geographical environment feature, and the feature of the wireless propagation path from the wireless cell to be predicted to the corresponding grid, the coverage indicator value of the grids using the trained coverage indicator prediction model, the method further comprises:
training the coverage indicator prediction model,
wherein training the coverage indicator prediction model comprises:
obtaining coverage indicator sample features and cell grid granularity label data of a wireless cell and a corresponding grid in the designated region; wherein the coverage indicator sample features comprise: a wireless cell feature of the wireless cell, a geographical environment feature of the wireless cell, a grid geographical environment feature, and a feature of a wireless propagation path from the wireless cell to the corresponding grid; and the cell grid granularity label data comprises: a statistical coverage indicator value of a same wireless cell in a same grid obtained by aggregating all coverage indicator values of the wireless cell in the grid to all terminals; and
training the coverage indicator prediction model according to the coverage indicator sample features and the cell grid granularity label data of the wireless cell and the corresponding grid in the designated region.
2. The method according to claim 1 , wherein obtaining the coverage indicator sample features and the cell grid granularity label data of the wireless cell and the corresponding grid in the designated region comprises:
obtaining wireless cell data for describing the wireless cell in the designated region, map data for describing a geographical environment in the designated region, and coverage signal data from the wireless cell in the designated region to a terminal;
constructing the cell grid granularity label data according to the coverage signal data; and
determining the coverage indicator sample features according to the wireless cell data for describing the wireless cell in the designated region and the map data for describing the geographical environment in the designated region.
3. The method according to claim 1 , wherein before training the coverage indicator prediction model according to the coverage indicator sample features and the cell grid granularity label data of the wireless cell and the corresponding grid in the designated region, the method further comprises:
cleaning and filtering all cell grid granularity label data in the designated region; and
training the coverage indicator prediction model according to the coverage indicator sample features and the cell grid granularity label data of the wireless cell and the corresponding grid in the designated region comprises:
training the coverage indicator prediction model according to the coverage indicator sample features and the cleaned and filtered cell grid granularity label data of the wireless cell and the corresponding grid in the designated region.
4. The method according to claim 3 , wherein cleaning and filtering all cell grid granularity label data in the designated region comprises:
deleting cell grid granularity label data meeting a preset condition;
wherein the preset condition comprises at least one of:
a distance between a grid and a wireless cell greater than a distance threshold;
a grid outside a horizontal lobe width of a wireless cell antenna; or
an amount of coverage signal data of a wireless cell in a grid less than or equal to an amount threshold.
5. The method according to claim 2 , wherein obtaining the statistical coverage indicator value of the same wireless cell in the same grid obtained by aggregating all coverage indicator values of the wireless cell in the grid to all terminals comprises:
averaging and aggregating all coverage indicator values of the same wireless cell in the same grid to all terminals, to obtain the statistical coverage indicator value of the wireless cell in the grid;
or, averaging and aggregating all coverage indicator values within a receiving threshold range of the same wireless cell in the same grid to all terminals, to obtain the statistical coverage indicator value of the wireless cell in the grid.
6. The method according to claim 1 , wherein obtaining the wireless cell feature of the wireless cell to be predicted, the geographical environment feature of the wireless cell to be predicted, the grid geographical environment feature, and the feature of the wireless propagation path from the wireless cell to be predicted to the corresponding grid comprises:
obtaining wireless cell data for describing the wireless cell to be predicted, and map data for describing a geographical environment in a coverage area of the wireless cell to be predicted; and
determining, according to the obtained wireless cell data and the obtained map data, the wireless cell feature of the wireless cell to be predicted, the geographical environment feature of the wireless cell to be predicted, the grid geographical environment feature, and the feature of the wireless propagation path from the wireless cell to be predicted to the corresponding grid.
7. The method according to claim 1 , wherein the coverage indicator prediction model comprises a gradient-boosted tree model.
8. An electronic device, comprising:
at least one processor; and
a storage device having at least one program stored thereon which, when executed by the at least one processor, causes the at least one processor to implement the coverage indicator prediction method according to claim 1 .
9. A non-transitory computer readable storage medium storing a computer program thereon which, when executed by a processor, causes the processor to implement the coverage indicator prediction method according to claim 1 .
10. A model training method, comprising:
obtaining coverage indicator sample features and cell grid granularity label data of a wireless cell and a corresponding grid in a designated region; wherein the coverage indicator sample features comprise: a wireless cell feature of the wireless cell, a geographical environment feature of the wireless cell, a grid geographical environment feature, and a feature of a wireless propagation path from the wireless cell to the corresponding grid; and the cell grid granularity label data comprises: a statistical coverage indicator value of a same wireless cell in a same grid obtained by aggregating all coverage indicator values of the wireless cell in the grid to all terminals; and wherein grids are obtained by dividing a designated region; and
training, according to the coverage indicator sample features and the cell grid granularity label data of the wireless cell and the corresponding grid in the designated region, a coverage indicator prediction model for predicting a coverage indicator value of the grids.
11. The method according to claim 10 , wherein obtaining the coverage indicator sample features and the cell grid granularity label data of the wireless cell and the corresponding grid in the designated region comprises:
obtaining wireless cell data for describing the wireless cell in the designated region, map data for describing a geographical environment in the designated region, and coverage signal data from the wireless cell in the designated region to a terminal;
constructing the cell grid granularity label data according to the coverage signal data; and
determining the coverage indicator sample features according to the wireless cell data for describing the wireless cell in the designated region and the map data for describing the geographical environment in the designated region.
12. The method according to claim 11 , wherein
before training the coverage indicator prediction model according to the coverage indicator sample features and the cell grid granularity label data of the wireless cell and the corresponding grid in the designated region, the method further comprises:
cleaning and filtering all cell grid granularity label data in the designated region; and
training the coverage indicator prediction model according to the coverage indicator sample features and the cell grid granularity label data of the wireless cell and the corresponding grid in the designated region comprises:
training the coverage indicator prediction model according to the coverage indicator sample features and the cleaned and filtered cell grid granularity label data of the wireless cell and the corresponding grid in the designated region.
13. The method according to claim 12 , wherein cleaning and filtering all cell grid granularity label data in the designated region comprises:
deleting cell grid granularity label data meeting a preset condition;
wherein the preset condition comprises at least one of:
a distance between a grid and a wireless cell greater than a distance threshold;
a grid outside a horizontal lobe width of a wireless cell antenna; or
an amount of coverage signal data of a wireless cell in a grid less than or equal to an amount threshold.
14. The method according to claim 10 , wherein obtaining the statistical coverage indicator value of the same wireless cell in the same grid obtained by aggregating all coverage indicator values of the wireless cell in the grid to all terminals comprises:
averaging and aggregating all coverage indicator values of the same wireless cell in the same grid to all terminals, to obtain the statistical coverage indicator value of the wireless cell in the grid;
or, averaging and aggregating all coverage indicator values within a receiving threshold range of the same wireless cell in the same grid to all terminals, to obtain the statistical coverage indicator value of the wireless cell in the grid.
15. The method according to claim 10 , wherein the coverage indicator prediction model comprises a gradient-boosted tree model.
16. An electronic device, comprising:
at least one processor; and
a storage device having at least one program stored thereon which, when executed by the at least one processor, causes the at least one processor to implement the model training method according to claim 10 .
17. A non-transitory computer readable storage medium storing a computer program thereon which, when executed by a processor, causes the processor to implement the model training method according to claim 10 .