Network commodity sorting method and device, equipment and medium

Abstract

The invention discloses a network commodity sorting method and device, equipment and a medium. The method comprises the steps of obtaining user behavior data and commodity exposure data, wherein the user behavior data comprises a purchase intention sequence and a historical click sequence; inputting the user behavior data and the commodity exposure data into a multi-target estimation model constructed based on a multi-task learning framework to obtain a corresponding commodity conversion rate, wherein the multi-target prediction model comprises a commodity click rate learning task and a commodity conversion rate learning task; sharing an embedded layer parameter and a partial attention parameter between the commodity click rate learning task and the commodity conversion rate learning task;inputting the historical click sequence and the commodity exposure data into a click rate estimation model to obtain a corresponding commodity click rate; determining the click conversion probabilityof the commodities based on the commodity conversion rate and the commodity click rate, and sorting the commodities based on the click conversion probability, wherein sorting can be conducted according to the comprehensive quality of the commodities, and then the commodity conversion rate and the total transaction volume are increased.

Description

technical field [0001] The invention relates to the field of online shopping, in particular to a method, device, equipment and medium for sorting online commodities. Background technique [0002] Currently, the click through rate (CTR, Click Through Rate) and conversion rate (CVR, Conversion Rate) of online shopping mall products are not completely proportional. For example, the cover design of a certain product is beautiful, and the click-through rate of the product is high, but the actual quality of the product does not conform to the description of the product, so users will not buy it after clicking; or the cover of a certain product is not attractive enough, and the probability of being clicked by users is low. Low, but the performance of other indicators such as the quality, evaluation and historical sales of the product is very good. Once the user clicks and enters the product details page and observes a series of historical performance of the product, the probability...

AI Technical Summary

Problems solved by technology

[0003] In the prior art, the conversion rate of the product is predicted by the model, but the conversion rate prediction model assumes that the product is clicked and predicts the probability of the product being converted, that is, the traditional conversion rate prediction model usually uses the click data as the training set, where the click but the Unconverted products are negative examples, clicked and converted products are positive examples, so there is a sample bias problem, which reduces the generalization ability of the model

In the prior art, the click through & conversion rate (ie, CTCVR, Click Through & Conversion Rate) of the product is directly predicted through the ESMM network structure. However, in practical applications, when the online sorting is performed according to the output CTCVR score, the click through rate will be large. The phenomenon of the magnitude of the decline reduces the click-through rate of the product, which in turn affects the conversion rate of the product and the total turnover

Images