Latent stage model for carsharing usage frequency estimation with Montreal case study

In order to predict the monthly usage frequency of members of a car-sharing scheme by analysing the gradual change of behaviour over time, a new model is proposed based on the Markov Chains model with latent stages. The model accounts for changing patterns of frequency from soon after signing up to later stages by including five latent user 'life stages'. In applying the model to panel data from Montreal's free-floating carsharing service the authors calculate each user's 'lifetime' applied to 'system operation time', the time period since the start of the scheme. Three-fold validation reveals effective performance of the model for both lifetime and system operation time dimensions. The model is further applied to illustrate how previous carsharing experience and the extension of the scheme to a larger area can affect usage frequency changes. We conclude that this approach is effective for usage prediction for novel transport schemes.

Automated summary

The study proposes a new model based on Markov Chains to predict the monthly usage frequency of members in a car-sharing scheme. By including five latent user 'life stages' and validating the model on panel data, the effectiveness of the model in predicting car-sharing usage frequency is demonstrated. This approach is effective for predicting usage in novel transport schemes.