Presentation is loading. Please wait.

Presentation is loading. Please wait.

TU/e Eindhoven University of Technology Exploring Heuristics Underlying Pedestrian Shopping Decision Processes An application of gene expression programming.

Similar presentations


Presentation on theme: "TU/e Eindhoven University of Technology Exploring Heuristics Underlying Pedestrian Shopping Decision Processes An application of gene expression programming."— Presentation transcript:

1 TU/e Eindhoven University of Technology Exploring Heuristics Underlying Pedestrian Shopping Decision Processes An application of gene expression programming Ph.D. candidateWei Zhu ProfessorHarry Timmermans

2 TU/e Department of architecture, building & planning Introduction  Modeling pedestrian behavior has concentrated on individual level  Decision processes only receive scant attention  As the core of DDSS, are current models appropriate?  Introducing a modeling platform, GEPAT  Comparing models of “go home” decision

3 TU/e Department of architecture, building & planning Random utility model  Discrete choice models have been dominantly used  Question 1: Too simple  Only choice behavior is modeled, ignoring other mental activities such as information search, learning  Question 2: Too complex  Perfect knowledge about choice options is assumed  Utility maximization is assumed  Degree of appropriateness?

4 TU/e Department of architecture, building & planning Heuristic model  Simple decision rules  E.g., one-reason decision, EBA, LEX, satificing  Human rationality is bounded, bounded rationality theory  Searching information—Stopping search—Deciding by heuristics  Degree of appropriateness?

5 TU/e Department of architecture, building & planning Difficulties in heuristic model  Implicit mental activities Test different models  Structurally more complicated Get simultaneous solutions  Irregular function landscape Effective, efficient numerical estimation algorithm Bettman, 1979

6 TU/e Department of architecture, building & planning The program--GEPAT  Gene Expression Programming as an Adaptive Toolbox  Gene expression programming (Candida Ferreira 2001) as the core estimation algorithm  Two features:  Get simultaneous solutions for inter-related functions  Model complex systems through organizing simple building blocks

7 TU/e Department of architecture, building & planning Genetic algorithm  GA is a computational algorithm analogous to the biological evolutionary process  It can search in a wide solutions space and find the good solution through exchanging information among solutions  It has been proven powerful for problems which are nonlinear, non-deterministic, hard to be optimized by analytical algorithms

8 TU/e Department of architecture, building & planning Get simultaneous solutions  The chromosome structure in GEP  Only one function can be estimated -b 2 +b+bd-c

9 TU/e Department of architecture, building & planning Get simultaneous solutions  The chromosome structure in GEPAT  Parallel functions can be estimated simultaneously.

10 TU/e Department of architecture, building & planning Test different models  Facilitate testing different models through organizing building blocks--“processors”  Each processor is a simple information processing node (mental operator) in charge of a specific task

11 TU/e Department of architecture, building & planning Parallel computing  Message Passing Interface (MPI)  Distribute computation by chromosome or record Master Slave

12 TU/e Department of architecture, building & planning Model comparison  Go home decision  Data: Wang Fujing Street, Beijing, China, 2004  Assumption: The pedestrian thought about whether to go home at every stop.  Observations: 2741 Shall I go home?

13 TU/e Department of architecture, building & planning Reason for going home  Which are difficult to observe  Using substitute factors  Relative time  Absolute time

14 TU/e Department of architecture, building & planning Time estimation  Estimate time based on spatial information  Grid space  Assumption  Preference on types of the street  Walking speed 1 m/s

15 TU/e Department of architecture, building & planning Multinomial logit model  Choice between shopping and going home Go home Shopping

16 TU/e Department of architecture, building & planning Hard cut-off model  Satisficing heuristic  Lower and higher cut-offs for RT and AT LC RT HC RT LC AT HC AT P NS Go home

17 TU/e Department of architecture, building & planning Soft cut-off model  Heterogeneity, taste variation LCM RT LCSD RT HCM RT HCSD RT LCM AT LCSD AT HCM AT HCSD AT P NS

18 TU/e Department of architecture, building & planning Hybrid model  When the decision is hard to be made, more complex rules are applied

19 TU/e Department of architecture, building & planning Model calibrations MNLHard Cut-offSoft Cut-offHybrid PValueP P P β1β1 -0.007LC RT 29.797LCM RT 132.048LCM RT 0.000 β2β2 -0.008--LCSD RT 83.976LCSD RT 327.290 β3β3 -10.501HC RT 674.966HCM RT 676.000HCM RT 676.992 ----HCSD RT 0.010HCSD RT 0.010 --LC AT 809.840LCM AT 927.851LCM AT 916.544 ----LCSD AT 87.422LCSD AT 85.820 --HC AT 1313.169HCM AT 1305.591HCM AT 1377.659 ----HCSD AT 104.161HCSD AT 230.719 --P hNS 0.308P hNS 0.752β1β1 -0.047 ------β2β2 0.000 ------β3β3 -3.502 ML-1121.200-1381.830-1070.599-1077.843 AIC2248.4002773.6602159.1992177.687 Sim0.5460.6560.7430.744

20 TU/e Department of architecture, building & planning Discussion  The satisficing heuristic fits the data better than the utility- maximizing rule, suggesting bounded rational behavior of pedestrians  Introducing the soft cut-off model is appropriate and effective; pedestrian behavior is heterogeneous  Lower cut-offs, as the baseline of decision, are much more effective than high cut-offs in explaining data, suggesting that pedestrians rarely put themselves to the limit in practice

21 TU/e Department of architecture, building & planning Future research  Model other behaviors, e.g., direction choice, store patronage, environmental learning  Compare models  Improve GEPAT

22 TU/e Eindhoven University of Technology Thank you Wei Zhu w.zhu@tue.nl Harry Timmermans h.j.p.timmermans@bwk.tue.nl


Download ppt "TU/e Eindhoven University of Technology Exploring Heuristics Underlying Pedestrian Shopping Decision Processes An application of gene expression programming."

Similar presentations


Ads by Google