Analyze the impact of the epidemic on New York taxis by machine learning algorithms and recommendations for optimal prediction algorithms

Abstract

With the growth of the economy and population, the transportation infrastructure of cities faces various challenges such as lacks of parking spots and traffic jams. One of the alternatives is to take public transport combined with proper utilization of demand-responsive transport such as Taxi or Uber. Therefore, it is valuable to improve the efficiency of these responsive transport. This research evaluates the predicting performance of different prediction models such as GBDT, XGBoost, and Random forest on taxi trip duration. At the same time, new elements were added to the modified data to contribute a higher accuracy rate including the snow status and precipitation data. Moreover, exploratory data analysis and data mining was conducted taking concerns of the Covid-19 effect of the year 2020.