I am working on my thesis (making the traffic lights system work more efficiently by letting them learn) and in my first part of this research, which is how to predict the traffic intensities of the next fifteen minutes, I have to predict the traffic (car) intensities for each lane of the intersection.
The figure above is a sketch of a real intersection and there are 12 different lanes of which I have historical intensity data. I want to use a regression model algorithm by predicting the intensities of each lane for the next 15 minutes.
The data set I have generated is a CSV-file containing all the lanes in the headers and the intensities for each lane in the Xth-15 minutes (3 months long). Below here I will show the generated data set.
I wrote a Console Application in C# to write and generate this output and I uploaded this data set on my Azure ML project.
When I want to train a model, I can only select a single column, which means I can only train one model for a lane. My question is, as I have read in another question, should I train a new model for each lane and save it, or can I somehow convert my data to a more efficient data set and eventually train it as one train model?