E-commerce platform connects sellers and buyers, with the ultimate goal of promoting transactions and pursuing higher sales. Improving the matching degree of people and goods is the key to promote transactions, and the key to improving the matching degree of people and goods is to "let buyers find the goods they want".

based on the theoretical model of "people, goods and fields", this paper attempts to establish a user portrait labeling system for e-commerce platforms.

The overall process of shopping on the platform can be summarized as: entering the platform, selecting goods and placing an order for payment.

this process can be abstracted as "people, goods and fields".

that is, how the user enters the page of what goods, completes the selection and places an order for payment.

people correspond to users, goods correspond to commodities, and markets correspond to channels.

user characteristics can be constructed from five dimensions: user's basic information, kinship, location information, social characteristics and consumption characteristics, which show what kind of person the user is.

commodity characteristics can be constructed from two dimensions: category preference and category behavior crowd.

1. Category preference

The category preference label is defined as the product category that users prefer on the platform.

2. Category Behavior Crowd

The category behavior crowd tag is defined as the user's interaction behavior in the specified product category and time period on the platform. What is popular is what the user has done to the goods.

there are three elements in category behavior population: specific category, behavior type and time range.

there are five types of behavior type I * * *, namely, search, click, collect, add purchase and purchase.

time range 1 * * * has five options, namely, last 7 days, last 15 days, last 3 days, last 9 days and last 18 days.

The people tagged by category behavior crowd are all people who have specified behaviors in a specified category within a specified time range.

channels refer to the channels through which users get in touch with detailed pages of goods, which can be divided into activity channels and content channels. Through the channel characteristics, people who prefer a certain content form and content channel can be delineated, so that goods can appear in this position in a targeted manner.

the label system of user's portrait is established through the theory of "people, goods and fields".