Uncertainty in Choice Behaviour Prediction: a Hybrid Approach to Combine Fuzziness and Randomness
G Circella, M Dell?Orco, D Sassanelli, Technical University of Bari, IT
We propose a hybrid model for the prediction of travelers' choices, which permits us to cope with the random and the fuzzy uncertainty at the same time.
The effects of uncertainty are difficult to be evaluated while analysing travelers? choice behaviour. Travelers? choices are commonly influenced by different dimensions of uncertainty. The variability of the alternative attributes is one of the main sources of uncertainty and risk; on the other hand, travelers? perception of the characteristics of travel options is not easy to be predicted. Even greater are the effects of uncertainty on transportation system users in Urban Areas, due to the interactions between public transport vehicles and private ones.
Nowadays, the mathematical frameworks commonly used to deal with uncertainty are mainly based on random utility models or on fuzzy inference. Although they assume quite different hypotheses on the definition of the alternative attributes, both approaches show some positive properties in the prediction of choice behavior; in particular, random utility models are useful when dealing with the variability of alternative attributes and fuzzy inference when modelling travelers? perception of the alternatives.
Even if these methodologies have been commonly used in the description of these different types of uncertainty, namely randomness and vagueness, very few experiences are reported on the efforts made to deal with these two types of uncertainty simultaneously. However, the need for a combination of the two approaches is arising: the interest for fuzzy human perception of the alternative attributes is seen to be as important as the stochastic investigation of occurring events.
In this work, we propose a hybrid model, especially designed to cope with the random and the vague uncertainty at the same time. The aim of the research is to implement a methodology which can be useful in the prediction of choice behaviour, when both types of uncertainty are involved in travelers? choice behaviour. The model is based on the assumption that both random and vague uncertainty affect choices. Thus, the alternative attributes are designed in a hybrid way, referring to a random variability, due to the objective randomness of travel variables, and to a fuzzy uncertainty, related to a lack of knowledge (or familiarity) in the perception of the alternatives. A hybrid utility function is used in the definition of choice predictions.
In order to verify the consistence of the model predictions, the proposed methodology is applied to a choice context, in which transportation system users are called on to make a choice between different public transport lines. An analysis of the weights of the two terms, randomness and fuzziness, is made with reference to the dependence of public transportation use on both types of uncertainty.
Association for European Transport