Nonlinear optimisation for itinerary planning

Abstract

Tourism has become a prominent global leisure activity. Tight budgets and pressing commitments have led to the rise in popularity of short trips to cities. As a result, tourists face the dilemma of selecting and ordering points of interest to visit, within the time available to them. Currently, no applications offer this service and travel agencies no longer provide the customisation that tourists may desire. This dissertation aims to provide this service through the development and deployment of the back-end for a mobile application that provides an itinerary for tourists visiting Valletta for a full or half day. The taboo search was implemented in Python to solve the optimisation problem of selecting highly rated points of interest, corresponding to user preference, whilst minimising the duration of travel between them. The implementation was further complicated due to the inclusion of constraints such as the user’s time budget, point of interest opening hours, restaurant recommendations for lunch breaks and dinner. With the aim to provide a mobile back-end for an online itinerary planner, the designed itinerary planning algorithm was deployed as a web application using Microsoft Azure. Monte Carlo simulations were executed and analysis shows that the produced itineraries recommended highly rated, user specific points of interest with a short duration of travel between them, within 1 second of execution. Results show that for both the day trip and short session, the mobile back-end is still able to give user specific itineraries within 3 seconds while running on Microsoft Azure. These results also hold true when the system is loaded.