Bayesian Neural Networks for Forecasting Restaurant Table Bookings

In the restaurant business, the amount of bookings and guests on a given day will vary greatly depending on multiple factors, both internal and external. Examples of internal factors may be events held at the restaurant such as parties, quiz nights or special offers. External factors may be things such as time of year, holidays, weather or sports events taking place in the area near the restaurant. Being able to predict the number of guests that are going to attend on any given day is very valuable to any restaurant owner as it lets them plan the amount of food and drinks to order and to schedule enough staff. This thesis aims to explore the possibilities of forecasting the number of restaurant bookings based on time of year and month in order to find patterns, both within a month and seasonal patterns across an entire year. To do so, a Bayesian neural network will be employed, learning patterns from previous booking data. Despite a limited number of factors, the results point in a positive direction towards further exploring the possibility of using such a network to forecast the number of restaurant bookings. By adding additional information and factors when making a forecast, this method could prove very useful for restaurant owners in the future by exploiting patterns in previous booking data