Research on Feature Engineering Training Model System for Deep Optimization of Merchant Transaction Data
Download as PDF
DOI: 10.23977/esac2022.012
Corresponding Author
Fenwei Guo
ABSTRACT
With the continuous increase of the number of enterprises, especially the continuous expansion of the scale of small and micro merchants, there are more and more risky merchants with illegal transactions and fraudulent behaviors. Based on the sales records of commodities and the basic information of merchants, this paper proposes a false transaction identification method combining the Stacking fusion model and multi-layer perceptron, and identifies false transactions by identifying commodities that can increase sales by brushing orders. First, the deep belief network is used to learn the transaction features to obtain higher-level abstract features; then the multi-layer perceptron is used to perform the classification task to identify fake transactions. The transaction records and comment data of products are crawled from Taobao for experimental verification. Compared with the experimental results of other machine learning models, its performance has been significantly improved.
KEYWORDS
Merchant Transaction Data, Fake Transaction, Deep Optimization, Feature Extraction, Stacking Fusion Model