CRNet: a multimodal deep convolutional neural network for customer revisit prediction.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Mobile food delivery services are vital to the restaurant industry.
• Predicting customer revisits is a significant research challenge.
• Multimodal datasets are increasingly used to solve complex industrial problems.

Purpose of the Study:

• To introduce CRNet, a novel multimodal deep convolutional neural network.
• To predict customer revisits in the context of mobile food delivery.
• To evaluate CRNet's performance against state-of-the-art models.

Main Methods:

• Developed CRNet, a multimodal deep convolutional neural network.
• Utilized two datasets: customer repurchase dataset (CRD) and mobile food delivery revisit dataset (MFDRD).
• Compared CRNet against two existing state-of-the-art multimodal deep learning models.

Main Results:

• CRNet achieved high accuracy: 0.9575 (CRD) and 0.9436 (MFDRD).
• CRNet achieved high F1-Scores: 0.9730 (CRD) and 0.9509 (MFDRD).
• CRNet significantly outperformed existing models (accuracy: 0.7417-0.9012; F1-Score: 0.7461-0.9378).

Conclusions:

• CRNet demonstrates superior performance in predicting customer revisits.
• The multimodal deep learning approach is effective for food delivery analytics.
• Future work can enhance the framework with additional data sources like metadata.