A Prediction Method of User Purchase Behavior Based on Bidirectional Long Short -Term Memory Neural Network Model
Haotian Wu · 2021
With the development of big data technology, predicting users’ purchasing intentions through historical data of users' purchasing behaviors has become an essential strategy for companies to carry out precision marketing and increase sales volume. The data of users’ purchasing behavior is characterized by a large amount, substantial variability, and potential long-term dependency. Therefore, the bidirectional long short-term memory neural network (BiLSTM) model is used in this paper to predict the user's purchasing behavior. First, the model takes user ID as the benchmark of the sequence, capturing the fluctuation law of the user purchase volume and fully mine the long-term dependency of user's purchasing behavior. Second, the BiLSTM model adaptively extracts features, realizes "end-to-end" prediction of user purchase behavior, and reduces feature engineering's subjectivity. This paper verifies the effectiveness of this method based on the real user purchasing behavior data sets. The experiment results prove that the BiLSTM method has high accuracy in the prediction of user purchase behavior.