Recommending items in pervasive scenarios: Models and experimental analysis