Spatio-temporal analysis of Uber NYC

Home Research Question About Us

Results

The research questions have been answered using different methods. You can find them in the chapter of 'Analysis' under choropleth map and animated map.

Research Question 1

Our analysis have shown that there clearly is a different behavior in the use of Uber depending on weekday and holiday. We have shown that the use of Uber is much higher on a usual weekday then on a usual holiday. Already the day before holiday has shown a slightly lower count of pick-ups but has not been as low as on the holiday itself. The animated point map indicates that the distribution of pick-ups over time differ for the two time-frames.The choropleth maps show the rate of change of the neighbourhood tabulation areas where Uber pick-ups occured. It shows very nicely that the highest rate of change were in areas where the number of pick-ups are very high, especially in Manhattan, whereas areas with low pick-up rates are less affected by the holiday. This analysis could not answer why there is a change in behavior and we can only speculate about the reasons for this. One could be that people left New York City and took occassion of a long weekend. Another reason could be that Uber is mostly used by commuters or by working people during the day. Furthermore the low rate of pick-ups could also be explained due to a low availability of Uber rides.

Research Question 2

A spatia-temporal analysis of the use of Uber can be done by different methods. For each method there are advantages and disadvantages. We could not find one method that fits clearly the best to answer research question 1. The use of a choropleth map allows a good overview on the change of pickups. Spatial patterns can be defined easily. Consequently it allows an easy comparison of the days. But the classification is not very clear, as the number of pick-ups varies very strong over space. Also the temporal variable over the day gets lost and the points of interests can not be defined. To compensate these disadvantages we suggest an animated map of points. This method allows to get insights on specific times and specific points of interests. Unfortunately, the comparison in this interactive map is much harder and the data processing is much slower. As a conclusion, there is no true or false on the choice of method and it is highly dependent on the purpose. For this project we have worked out two methods with different properties such as the combination can give a deeper insight into the behavior of the use of Uber.

Related Work

Spatio-temporal Analysis of Uber-Pickups in NCY

Since the invention of ridesourcing around 2009, the count of users and drivers has increased immense all over the world. As the phenomena is rather new, there are not yet many insights of the behaviours and possible development. Nevertheless, such analysis is getting more and more important with the increasing numbers in term of future planning transit systems.

Information associated with taxi trips can provide unprecedented insight into many different aspects of city life, from economic activity and human behaviour to mobility patterns. But the data complex, containing geographical and temporal components in addition to multiple variables associated with each trip. Consequently, it is hard to specify exploratory queries and to perform comparative analyses (e.g., compare different regions over time. In cities where taxis play an important role in urban transportation, investigations about taxi trajectory can be used to study intra-urban human mobility. In this context, models can be built where mobility patterns are shown in a spatio-temporal context (Liu et al. 2012; Ferreira et al. 2013). Ferreira et al. (2013) propose a model for analysing taxi routes. Their model can express a wide range of spatio-temporal queries, which can be applied on different aggregations and visually represented. The goal of their work is to unify the two phases of exploration: data selection and visual analysis. This allows user to explore and compare results. Liu et al. (2012) collected GPS-derived data of 6'600 taxis over on week. By analysing the taxis' trajectories, they find that globally spatio-temporal patterns of trips exhibit a significant daily regularity thus they examine a dominant trajectory direction within the study area. Pan et al. (2013) found by analysing pick-up and drop-off dynamics of taxi GPS traces that they exhibit clear patterns corresponding to the land-use classes of the region. Furthermore, they compute land-use classification based on taxi GPS traces (Pan et al. 2013).
Analysis and studies about Uber are rather new and some are determining the same datasets. The dataset of Uber Pickups in New York city has been determined in different manners. In an economic perspective, the question if Uber is used as a substitution or complement for public transport. Through aggregation of the Uber Pickups, concerning demographic relation and infrastructure, Hoffmann et al. (2016) present insights that Uber is being used as a short-run replacement for public transportation. According to Rayle et al. (2014) 67% of all trips by taxi or ridesourcing are of a sociale/leisure purpose (bar, restaurant, concert, visits). Only 16% of the rides have been used in term of work. Also, they conclude that 40% of the origins are home-based and almost half of all rides occurred on Friday or Saturday. This study was done on an empirical dataset where people have been questioned on the street within specific districts. Another geo-spatial analysis of two competing taxi companies (Uber vs. Green cab) shows that the demand for Green cab's is still growing, but the growing number of Uber rides in the same area is growing more rapidly. Additionally, the analysis showed that especially in relatively poor neighbourhoods, Green cabs are performing better than Uber's (Poulsen et al. 2016). In a different study the Uber pick-up dataset was analysed in term of spatial patterns. A heatmap shows the distribution of users over each day and on the weekend. A crucial step using this dataset is aggregation and segregation (Kumar et al. 2014). Because as a lot of other location-based application, Uber captures massive amounts of data. This immense count of data, the so named big data, brings some challenges and it is important to be aware of them. To do efficient analysis it is helpful to think about different ways of geospatial data mining such as clustering or classification (Pandey et al. 2016). This also leads to many opportunities in visualising the data in different ways using different clustering and aggregations (Van Wijk 1999). However, researchers should find approaches to deal with complexities of current data to support spatio-temporal thinking and contribute to solving a large range of problems (Andrienko et al. 2010).