I'm running a pizza service and would like to predict a number of orders for every hour interval during a day per location (basically where we should deliver the pizza) in future.
And I've got a small data set of past orders: pairs of [timestamp, location], where location is a pair of latitude and longitude.
I did a small research and they say the simplest way to predict is to use a linear regression model with multiple variables (3 in my case: *timestamp, lat, lon).
So I've got NUMBER_OF_ORDERS(TIME, LATITUDE, LONGITUDE) function. Does it make sense to use a linear regression model for my task? I'm interested in the most simple yet accurate solution.
Linear regression finds a line of best fit between a dependent variable and one or more independent variables. Linear regression only makes sense if you believe there could be a linear relationship between the dependent variables and the independent variables. I can't see why number of pizza orders should be a linear function of either time or locations, so I doubt linear regression will give you any useful information.
But I believe you have a more fundamental issue. I don't think you have enough data to build any sort of meaningful model. Number of pizza orders must depend significantly on other factors apart from time of day and delivery location. Day of the week must be a big factor, and I imagine there are seasonal trends as well.
In other words, knowing that there are an average of 10 pizza orders between 2pm and 3pm doesn't help you if the number on any given day varies from 2 orders on a Monday in January to 20 orders on a Saturday in July.