| Summary: | 碩士 === 國立成功大學 === 交通管理(科學)學系 === 85 === The purpose of this research is to discuss the factors that
affect the updating effect of mode-choice models. Using
multinomial logit model, I compared and analyzed the
updating effect of several issues, i.e., updating methods,
sample size, utility function specification, data type, and
alternative choice set. The intercity trip data between Tainan
to Taipei werecollected for the empirical study. The empirical
results showed:
1. Among the four updating methods, transfer scaling and joint
context estimation had the best updating effect.
However, the easier updating procedure of transfer
scaling made it an effective and efficient updating method.
2. In the case of same alternative choice set, the sample size
of 120 was large enough for model updating. When the
updating samples used was 180, reestimating the model would
yield better results than updating models. However, in the
case of different alternative choice set, the sample size
must be increased to 180 to get a good updating result.
3. The performance of utility function specification had great
influence on the updating models. The specification effect
was greater than updating effect. The travel behavior
revealed from the exponential value of travel variables
showed that the marginal utility of travel cost was diminishing
while travel time was increasing.
4. The best updating combination between two data sets was
original RP data plus updating RP or JP data. The ratio of
1:2 between RP and SP for the JP data worked well.
5. The updating methods could apply to added new alternative
mode situation. However, the underlying assumption of IIA
must be satisfied.
|
|---|
