Assume we start on one huge area and want to build a model to suggest whether how much for each point of latlon recommended to be placed with new signal transmitter.
1st step : create all possible latlon within border area (100m grid)
2nd step : create all relevant features like
-closest existing transmitter distance -current signal level for the point (at this sample latlon) -how much use using the existing transmitter that closest to this point (at this sample latlon) -number of users using signal at this current point
Questions
1.If I choose NNet model, I need to come up with score function to calculate suitable score for each feature vector. because it is required to train with the model since its supervised model, how can I find the suitable score function? and do I need to consider outlier at particular feature when find the score function.
2.If I can have labeled data, what is the benefit of ML model since I can use this score function with automated traditional software to generate the rank of point list.
3.If I use unsupervised method, it will group only similar characteristic point based on features, but how to rank this point, somehow, I need to educate model that each feature, the higher value means good, or higher value means bad right?
4.any recommendation or any better model choice?
I have tried to outline toe step, but I have question since the approach does not sound to outperform traditional automated software.
Top comments (0)