Geography Reference
In-Depth Information
We consider 'age' as a representative of socio-demographic variables because it
is a proxy for employment status, life cycle and life style. It is thus an important
explanatory variable for activity patterns. Further, activity patterns on weekdays and
weekends are usually quite different, thus day of the week in terms of the division
between weekdays and weekends is also included.
If all five variables are used to encode the chromosome, the algorithm will
become very complicated. Instead we stratify the cases by day of the week and
age of respondents into four cases: younger or older than 25 years old and weekday
or weekend. The other three variables are used as genes to develop chromosomes
specifically for each of the four cases. The Michigan approach is used to encode the
chromosomes. In other words, separate classification models are developed for each
purpose. The encoded chromosome is illustrated as follows:
if ( C
j
∈) and (V
1L
A
1
) and (V
2L
A
2
V
2R
), then C
j
.
Conditions
Conclusion
The first condition (C
j
2
) is to judge whether C
j
is one of the possible trip
purposes; A
1
and A2 are the attributes representing start time and duration for
stay respectively, V
iL
,V
iR
are left and right boundary value of attributes; C
j
represents one of the trip purposes. In this study, seven categories of trip purpose
are differentiated including home (C1), work (C2), school (C3), shopping (C4),
recreation (C5), personal affairs (C6), and volunteering/religion activities (C7).
Following Freitas (
2002
), the fitness function is defined based on the confidence
factor and the completeness factor. In other words, the fitness value of each chro-
mosome depends on how well the attributes match with those existing combination
of conditions and conclusion derived from data provided by respondents.
In order to ensure that stronger chromosomes have greater chance of survival
than the weaker ones, firstly we use the elite strategy to reserve the best individuals
in each generation. In other words, the chromosome with the highest fitness value in
the current generation directly survives into the next generation without crossover or
mutation. The other chromosomes are selected through the roulette wheel selection
process, which ensure that each chromosome has the chance to be chosen and the
ones with higher fitness values are more likely to be selected.
Chromosomes selected from the above method are then randomly chosen to
be operated by one-point crossover through crossover rate (a threshold between 0
and 1 to judge which chromosomes should conduct crossover operation). In case
unreasonable chromosomes are generated during the process, this step would be
operated to the boundary values of the same attribute like starting time or duration
from different individuals. Similarly, mutation also operates to the chromosomes
which are randomly chosen through mutation rate (a threshold between 0 and 1 to
judge which individuals should conduct mutation operation), and we can perform
this operation by changing boundary value of one attribute to another random
number between 0 and 1.
Search WWH ::
Custom Search