Trip Matrix Level Analysis

The last measure to evaluate the predictive performance between CHAID and BNT is calculated at trip level. The origins and destinations of each trip, derived from the activity patterns, are used to build OD-matrices. The origin locations are

Table 7.4: SAM distance measures for activity pattern level analysis.

Table 7.5: Correlation coefficients betweeb OD-matrices for trip level analysis.

represented in the rows of the matrix and the destination locations in the columns. The number of trips from a certain origin to a certain location is used as a matrix entry. A third dimension was added to the matrix to break down the interactions according to some third variable. The third dimensions considered are day of the week, transport mode and primary activity. The bi-dimensional case (no third dimension) was considered as well. In order to determine the degree of correspondence between predicted and observed matrices, a correlation coefficient was calculated. To this end, cells of the matrix were rearranged into one array and the calculation of the correlation is based on comparing the corresponding elements of the predicted and the observed array. The results are presented in Table 7.5.

It can be seen from Table 7.5 that the good performance of BNT at the activity pattern level could not be maintained at trip level. The correlation coefficient is especially low for the OD matrix, where the primary activity is taken as the third dimension. Although this should be the subject of additional and future research, it is believed that the integrated approach did predict less activities in the activity schedule than CHAID. This deficiency is less apparent at the pattern level than at the trip level.