Análise de intervenção em séries temporais: aplicações em transporte urbano

Abstract

The present work aimed chiefly to analyse the behavior of urban transportations in the city of São Paulo by using intervention analysis in time series. Currently, 54 companies operate with a fleet ofabout 10 thousand units, 800 transport lines utilized by almost 4 million passangers a day. Three series were considered; lhe fírst reports the mean daily number of passangers, encompassing the period from January of1983to December of 1998 with 192 month averages; the second fumishes the number ofholdups in urban buses with measurements from Januaryof 1992to August of 1999, amounting to 93 month observations; and last series reports the number oftraffic accidents with buses over the period from January of 1988to September of 1999, with 141 month observations. Box and Jenkins models were fitted, with additions parameters to account for intervention. Identification of models was accomplished fimction autocorrelation and function autocorrelation paitial for the stationary series. Reasonability offitted models was evaluated by Box and Pierce test. Amongst reasonable modelsfor eachseries, onewas chosen according to Akaike criterion and residual mean square values. The last 6 observations of each series were disregarded fromlhe analysis to be compared withthe forecasted values ofeach model. Some events showed evidence signifícative of intervention, as the increase of the number of transport lines in the city of São Paulo and the privatization ofC.M.T.C. in 1993.