Identifying purchase intention through deep learning: analyzing the Q &D text of an E-Commerce platform

Identifying purchase intention by analyzing the Query and the Document of the product description (Q &D) text is one of the most important means of promoting Purchase Rate (PR). In view of that customers sometimes cannot describe their purchasing intention in queries, this paper aims to identify purchase intention from implicit queries by computing semantic similarity between Q &D and proposes a novel model based on Word2Vec algorithm, Long Short-term Memory (LSTM) and Deep Structured Semantic Model (DSSM). Besides, an empirical analysis is conducted through the Keras framework and based on the factual retrieval data of the Home Depot, an E-commerce website selling building materials in America. The results show that the proposed model has achieved improving F1-score on test dataset compared with other existing models. The novel model combines Word2Vec and LSTM to extract text features and applies DSSM to further fetch high-dimension representations by maximizing semantic similarity between the user query and the description of the correct merchandise. Our proposed model can be used to remove or minimize subjective factors in extracting features, improves the performance of purchasing intention identification, and also improves the customer experience of online shopping.