A comparison of learning methods over raw data: forecasting cab services market share in New York City

摘要

The cab services, present in most of the cities, are one of the most used offerings for passenger transportation. Nowadays their business model is being threatened by the meddling of emerging third parties powered by modern technologies. Based on the New York cab data, we will make a comparison of several machine learning techniques (linear regression, support vector machines and random forest) for forecasting the amount of dollars spent in the cab service. The comparison of those methods will focus on the accuracy of their forecasts under several circumstances: real data applied to all features, some noisy data (real data with some uniform distributed noise added) applied to several key features and some estimated data (obtained from other statistical estimators) applied to the key features. The main goal of this comparison is to provide some data regarding the performance of those methods when they are used in conjunction with other estimators