2.6 Conclusion

In summary, phase 2 of the project did not find sufficient evidence to support the claim that Total Amount is higher for trips starting and ending in hotspot areas, as only the airport areas follow such trend. However, Manhattan trips are more consistent in their price range, and coupled with our knowledge from phase 1 findings about the overwhelmingly high demand of taxi in this area, there is no reason for taxi drivers not to target the core of Big Apple as it will yield consistent income despite a low per-trip amount. The modelling also supports the correlation between lower accessibility to public transport and taxi price, implying that Manhattan areas that are not directly adjacent to subway are potentially great pick-up areas for taxi drivers. These analyses constitute a first step toward quantifying the relationship between factors affecting taxi profitability using a self-explanatory glass-box model in place of complex black-box models.

Aside from minor recommendations throughout the analysis, a major drawback of the analysis is the exclusion of the time and zone attributes from the model due to limited resources. Future work should investigate the relationship using a time-dependent model such as a variation of the LME introduced in this report with a more refined resolution of temporal and spatial structure to further reduce the error range, effectively improving the strength of the inferences.