I am preparing the Bus movement dataset for deep learning (ANN/CNN/RNN) analysis for congestion events detection. This is an extension to my original question, which can be located at 'Deep learning model training and processing requirement for Traffic data' for the general approach on this topic, and this question is for preparing the dataset and need your kind advice on it. In simple words, I would like to know the state of congestion for a bus route at a specific point in time (year).
Here are my entities:
- Routes
- Bus_Scheduled_Routes
- Bus_Route_Stops
- Bus_Trips (operational_date, Vehicle_id, Trip_id, Vehicle_Position_Update, Trip_stop_id, passenger_loaded, velocity, direction, scheduled_arrival_time, actual_arrival_time)
- Events (human and non-human induced)
- Points of Interests (POIs)
If I have these entities based data and I create a view that gives me a time reference based view comprising of week(52), day(7), Vehicle_id, Trip_id, Stop/Position_update_interval, speed, acceleration, velocity, scheduled_arrival_time, actual_arrival_time. Will this view be recommended to start training the model?
Secondly, how can I integrate the human / non-human induced events and Points of Interests (POIs) data into this view so my model can predict better results? To generalize the model data will be 'time segment / trips time (seasons), location component (Bus Routes and Stops), Arrival time / trip completion time'. I am thinking to add an attribute for human/non-human induced events as type tying with the 'time segment' and adding the POIs as type and vicinity to the stop points. What are your recommendation about it? Thanks in advance for your help.
