Activity-based model based on long short-term memory network and mobile phone signalling data

Abstract

With the advent of big data era, activity-based model (ABM) has once again become hot topics in the traffic planning. Traffic big data can reflect individual travel patterns, making it possible to establish ABMs. However, current ABMs based on big data are not mature, especially in the individual trip forecasting. Therefore, this paper proposes an advanced ABM using Long Short-Term Memory (LSTM) networks and mobile phone signalling data. The model is skeleton scheduling which contains primary activity chaining and secondary activity nesting. Then a time-dynamic adjustment model is proposed to adjust time conflicts among consecutive activities. A field test is conducted in Chengdu. The KS values of work and leisure departure time reach 35.20 × 10⁻² and 41.02 × 10⁻² separately, and that for activity duration reach 44.91 × 10⁻² and 54.65 × 10⁻². The results show our model can effectively predict activities, and has better accuracy and stability than existing BN, DT, GRNN, RF and GRU.

KEYWORDS:

• Activity-based model
• mobile phone signalling data
• LSTM
• skeleton schedule
• travel preference

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by National Natural Science Foundation of China: [Grant Number 52072313, 52002030]; Humanities and Social Sciences Foundation of Shannxi Province: [Grant Number 2020R035]; Natural Science Foundation of Shannxi Province: [Grant Number 2021JQ-256]; Humanities and Social Sciences Foundation of the Ministry of Education: [Grant Number 20XJCZH011]; Fundamental Research Funds for the Central Universities CHD: [Grant Number 300102341676, 300102342105].