How to train a model to predict the number of people at a certain bus stop before they cumulate in large numbers?

Question

Each person probably uses an app that tracks his/her position periodically and sends it to our servers. What I want is to use these data to train a model to predict the rush hours of each bus-stop on the map, so we can send extra buses to handle the predicted cumulation before it happens.

I have no experience in AI nor machine learning. So, which model should I use to do this?

Answer 1

Maybe you can use a recurrent neural network on saved data to train a predictive model based on past data.

Answer 2

H2O's AutoML is the thing that you are looking for believe me it will make your life super easy.

So how it generally works is:

• You have data and want to make sense out of it by making a prediction/classification and whole other array of things, in your case prediction.
• Suppose there are 50 prediction algorithms out there, it's highly unlikely that an individual knows all of them. That's where AutoML comes into the picture.
• You give AutoML some part of your processed data and tell it to find a perfect algorithm that he thinks you should use on this data for this type of prediction. See the usage of the AutoML's API in the doc and videos on youtube.
• It then gives you a list of best algorithms that AutoML thinks is best based on the loss function that you specify. There are many other parameters that you can specify in the AutoML's API.
• Pick top 1-2 and tune the hyperparameters of the algorithm.
• That's it.

Even then if you are not happy with the performance of your system try out ensemble learning before you jump for the power of neural nets which comes with lots of complexities and performance issues.