Cross-platform e-commerce fraud detection method and system based on comment data

Abstract

The invention discloses a cross-platform e-commerce fraud detection method based on comment data, comprises the following steps: obtaining comment data of goods from relevant e-commerce websites, manually labeling commodity attributes corresponding to the comment data, extracting word level features, commenting semantic features and comment structure features, and constructing a training set; taking the extracted features as input, the binary classifier being trained by using the training set; obtaining the comment data of the target goods from the relevant e-commerce websites, extracting thefeatures of the target goods, and inputting them to the trained binary classifier to identify the attributes of the target goods. The invention also discloses a cross-platform e-commerce fraud detection system. The detection method of the invention extracts platform-independent commodity features from three aspects of vocabulary, semantics and structure of e-commerce comments, and judges whether the commodity is suspected of fraud based on the features, so that the detection result is more accurate.

Description

technical field [0001] The invention relates to the technical field of e-commerce big data mining, in particular to a method and system for detecting fraudulent behaviors in cross-platform e-commerce based on review data. Background technique [0002] Today, e-commerce has become an efficient link between consumers, factories and retailers, providing consumers with a fast, convenient and reliable shopping environment. The many advantages of e-commerce have led to more and more consumers tending to shop online, making e-commerce flourish, and e-commerce retail sales have also grown rapidly, which has brought huge benefits to factories, retailers and e-commerce service providers. economic benefits. For example, Alibaba’s annual report shows that the total transaction volume of its e-commerce platform Taobao in 2017 reached 2,202 billion yuan; Amazon’s annual report shows that the total transaction volume of its e-commerce platform in 2016 reached 970 billion yuan; The total ...

AI Technical Summary

Problems solved by technology

[0005] For e-commerce service providers, they are limited by privacy protection and ethical issues to some extent, and cannot detect fraudulent products well

Even if some e-commerce service providers are willing to actively and responsibly maintain a benign e-commerce environment, the competition among e-commerce service providers makes it impossible for these service providers to cooperate with each other, and it is difficult to detect fraudulent products outside the e-commerce platform

Exploring fraud detection becomes more difficult when e-commerce internal data (e.g., user click data and user-item correlation graph) are not available

Therefore, some existing methods in academia cannot be directly applied to e-commerce fraud detection.

For example, when internal click data is not available, malicious click detection methods based on user clicks cannot be directly used for e-commerce fraud detection

Images