Aplicación del análisis mediante mapas auto-organizados (SOM) para estimar el uso de la bicicleta compartida: una nueva perspectiva

Abstract

The weather could be the key to forecasting whether the use

of active transport (depending on the city studied), but to

date, the predictions made have generated some controversy

for not being analysed by non-linear analysis. The objective of

this study was to examine the relationship between the time

spent using the Valencia bike sharing service (BSS) as a means

of active transport and the weather. A self-organizing map

analysis (SOM) was performed to generate profiles (clusters) of

the days of BSS use and current weather factors, plus a non

parametric analysis to compare the different profiles generated.

The results produced 8-day profiles with multiple significant

differences that showed that, although some variables have

greater weight than others in estimating BSS use, their

relationship is not always linear and a combination is needed

for more accurate predictions. We found that to predict high

use days, the weather should be warm with low to moderate

humidity, although temperature is limited if humidity is high,

with virtually no precipitation and a low average wind speed.

To estimate days of low BSS use, there should be both high

relative humidity, precipitation and wind speed. If the humidity

is not high and there is no precipitation on these days, low

temperatures can be taken into account. The results indicated

that the use of non-linear analyses such as SOM is an effective

tool for estimating BSS use in relation to the current weather

conditions.