Predicting displacement of passengers due to London Underground disruptions and alternative route generation.

KEY OBJECTIVES/AIMS OF THE RESEARCH:
INVESTIGATING RELATIONSHIPS/USER MODELLING
To develop an accurate model, the first objective is to investigate the relationship between the displacement of passengers and the different types of disruptions and model the users' behaviour. The aim is to understand how people may behave or react when there is unplanned disruption(s) on their journey.
PREDICTIVE MODEL
The main objective of the project is to develop a machine learning model to predict the displacement of passengers due to train disruptions on the London Underground network. The aim is to enable TfL officials and passengers to utilise the output of the prediction to minimise the disruption and make the journey run smoothly. TfL officials may use the predictions to deploy resource more efficiently and monitor the effects in real time. Passengers may use the prediction to plan an alternative journey that avoid much of the inconvenience.
PRESENTATION
The tertiary objective is to present the output of the prediction effectively, by using a combination of visualisation and or verbalisation as necessary. The aim is for the user (TfL officials and passengers) to make an informed decision based on the prediction they are presented with. Presenting the right type and amount of information is key for the parties to make their decisions quickly.
NOVEL METHODOLOGY:
Establish baseline state using historical data. The change in state (performance indicator) will be measured using Volume Over Capacity (VOC), Absolute, and Relative Difference Ratio (ADR and RDR).
Disruption scenarios will be simulated to identify and understand passenger behaviour by measuring the VOC, ADR and RDR.
Machine learning will be applied to learn the behaviour and predict displacement of passengers in a disruption state.
Genetic Algorithm to generate alternative routes to optimise network load and passenger experience.