Publication: Sistema recomendador de taxis para Big Data
Loading...
Identifiers
Publication date
2017-09-20
Defense date
2017-09-29
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Este documento propone un sistema para maximizar los ingresos de los taxistas,
recomendando las zonas de la ciudad con mayor frecuencia de viajes y más cercanas a la
posición del taxista en ese momento.
Analizando 10M de trazas de viajes realizados por los Taxis Amarillos de Nueva York
con un cluster Spark, encontramos correlaciones entre los beneficios y el tiempo,
distancia y número de pasajeros. También se encuentran que los lugares más frecuencia
de viajes cambian cada hora y cada día.
Mediante clustering, el sistema computa las agrupaciones más lucrativas para cada hora y
día de la semana, dando una puntuación a cada uno de los grupos basado en las
correlaciones encontradas. El sistema se ejecuta varias veces sobre un clúster Spark,
buscando la configuración más óptima.
Los resultados se guardan en una base de datos geoespacial, y puede consultarse
mediante una aplicación web introduciendo la hora, día de la semana, y ubicación. El
sistema recomienda las diez ubicaciones más cercanas, ordenadas por beneficio.
El sistema puede ser interesante para los Taxistas Amarillos de Nueva York, como una
forma de incrementar los beneficios influyendo en sus desplazamientos de forma directo,
libertad que los servicios competidores como Uber y Cabify no ofrecen, ya que los
objetivos de tales taxistas les son fijados por la compañía.
This document proposes a system to maximize the income of taxi drivers, recommending the areas of the city with more frequency of trips and closer to the position of the taxi driver at the time of the query. Analyzing 10M traces of trips made by the New York Yellow Taxis with a Spark cluster, we found correlations between the benefits and the time, distance and number of passengers. We also found that more frequent travel places change every hour and every day. Through clustering, the system computes the most profitable groups for each hour and day of the week, giving a score to each of the groups based on the correlations found. The system runs several times on a Spark cluster, looking for the most optimal configuration. The results are stored in a geospatial database, and can be viewed through a web application by entering the time, day of the week, and location. The system recommends the top ten closest locations, sorted by profit. The system may be of interest to the Yellow Taxi drivers in New York, as a way to increase profits by influencing their wandering, something that competing services like Uber and Cabify do not offer, since the objectives of such taxi drivers are fixed by the company.
This document proposes a system to maximize the income of taxi drivers, recommending the areas of the city with more frequency of trips and closer to the position of the taxi driver at the time of the query. Analyzing 10M traces of trips made by the New York Yellow Taxis with a Spark cluster, we found correlations between the benefits and the time, distance and number of passengers. We also found that more frequent travel places change every hour and every day. Through clustering, the system computes the most profitable groups for each hour and day of the week, giving a score to each of the groups based on the correlations found. The system runs several times on a Spark cluster, looking for the most optimal configuration. The results are stored in a geospatial database, and can be viewed through a web application by entering the time, day of the week, and location. The system recommends the top ten closest locations, sorted by profit. The system may be of interest to the Yellow Taxi drivers in New York, as a way to increase profits by influencing their wandering, something that competing services like Uber and Cabify do not offer, since the objectives of such taxi drivers are fixed by the company.
Description
Keywords
Big Data, Cluster, Data mining, Spark, Spatial clustering, Taxi