A multi-attribute decision making approach based on information extraction for real estate buyer profiling

With the rapid development of the Internet and the widespread usage of mobile terminals, data-driven user profiling has become possible. User profiles describe the user’s overall behavior characteristic from multiple perspectives (e.g. basic information, feature preference, social attribute), which can explore the potential relationships between complex user behaviors and the decision-making process. In this paper, we focus on the problem of real estate buyer profiling and propose a novel multi-attribute decision making (MADM) approach, trying to solve the needs of enterprises to locate target customers accurately. Firstly, we reorganize the dataset by integrating structured with unstructured data, where an Enriched Bi-directional long short-term memory (Bi-LSTM) Conditional Random Field (EB-CRF) model is proposed to extract important information in the unstructured data. Based on four general dimensions (i.e. basic information, family situation, purchase intention, financial situation), we then design an entropy-based weight allocation algorithm to obtain attribute weights, which helps explore implicit heterogeneous relationships. Finally, with the help of expert knowledge, we use attribute weights and representation technology “bag of attributes” to construct a buyer-specific feature representation. Extensive experimental results indicate that our approach outperforms strong baselines significantly and achieves state-of-the-art performance.