Combining Mobile Telephone and Railway Data : An Analysis Perspective

Abstract

The advancement of digital data and its accessibility brings about an unprecedented opportunity for its usage in transport planning. Mobile telephone data represents a major portion of population and is an attractive entity for estimation of user trends in a variety of public arenas. The usage of this type of data in benefit analysis of major transport projects is still unexplored. Mobile data has the potential to predict a variety of transport patterns and matrices. Number of travelers on a train is a confidential data and the approaches for its calculation have question marks over its reliability. The possibility of estimating this information by utilizing anonymous mobile data is an attractive prospect and has the possibility of cross validation for train operators and enhancing the accuracy of benefit analysis for railway infrastructure projects. The mobile data is combined with a specific type of transport data i.e. punctuality data for train travel. The thesis is focused upon revealing the multi-dimensional analysis possibilities. Initially the present data is subjected to reveal its visual and descriptive aspects. The analysis is kicked off with establishing the fact that the mobile subscribers on the trains passing adjacent to the base stations are indeed reflected in the count of mobile events and there is strong positive correlation between the two. This is followed by introducing the time and direction aspect to the train passing and test whether it is possible to verify as well as disclose trends related to the count of travelling passengers. The analysis is comprised of a systematic evaluation approach based on an algorithm to categorize the data into specific categories. The statistical significance of the different categories unveils the different patterns of effect of train data on mobile data. The Automatic passenger count provides the opportunity to add a comparative dimension to the analysis, the relation between is verified and quantified by calculating the ratio between the two for specific days, hours and train directions. The analysis reveals that the ratio is constant for specific timeframes. Prediction and Forecasting is one of the most pivotal dimensions of Big data potential. Accordingly, by developing mathematical models and forecasting future events the ability of the available mobile data for predictive analysis is ascertained. In addition to the background concepts and problem formulation, the thesis explains the analysis approaches, results and the inferences from these results in detail.