Insurance product recommendation method and insurance product recommendation system

Abstract

The invention provides an insurance product recommendation method and an insurance product recommendation system. The insurance product recommendation method comprises steps that a label of an insurance product is extracted, and is disassembled into various hierarchical labels of different levels, which are stored in a label database, and the labels are the characteristic attribute of the insurance product; according to data in the database used for storing the hierarchical labels, a label corresponding to a target product newly added to the system is generated; according to user behaviors corresponding to different users, the product labels are recommended for the users, and matching products are recommended indirectly. The users are matched with the labels, and the most adequate insurance product combinations are recommended for the users; because of the effective matching between the products, the guarantee effect is greatly enhanced; by adopting an expression way of disassembling the insurance product into components, the system is capable of recommending the insurance products, which are close to the reality, more accurately; by knowing different user groups preferring which labels and which product characteristics, the products, which are more adequate, can be recommended for the users by aiming at the preference of the different users.

Description

technical field [0001] The invention relates to the field of product recommendation systems, in particular to an insurance product recommendation method and system. Background technique [0002] First, the existing insurance product recommendation system is pushed to users based on the characteristics of the insurance product itself, but the components of each insurance product, such as insurance liability, insurance premium rate, premium payment method, insurance period, insurance compensation or insurance money The payment methods of insurance companies are very different, and it is difficult to recommend them to customers based on the two dimensions of good / bad, because an insurance can contain many features, so it is not known which feature attracts users. When a user chooses a certain product, the technical background cannot clearly identify which constituent element the user is optimistic about, or which element combination of the entire insurance product the user is o...

AI Technical Summary

Problems solved by technology

[0002] First, the existing insurance product recommendation system is pushed to users based on the characteristics of the insurance product itself, but the components of each insurance product, such as insurance liability, insurance premium rate, premium payment method, insurance period, insurance compensation or insurance money The payment methods of insurance companies are very different, and it is difficult to recommend them to customers based on the two dimensions of good and bad, because an insurance can contain many features, so I don’t know which feature attracts users

When a user chooses a certain product, the technical background cannot clearly identify which constituent element the user is optimistic about, or which element combination of the entire insurance product the user is optimistic about. Therefore, the user's needs cannot be recorded and analyzed by the background, which does not meet the demand-oriented model. insurance market requirements

[0003] The existing recommendation system can recommend products that customers should / may / generally like, using statistical tools such as demographics, but the specific accuracy rate is not high, because insurance products are composed of various components, mechanical The noise of the application combination will be very loud, and it is difficult to touch the core needs of customers

[0004] Second, there are thousands of life insurance products on the market, but there are only a few hundred products that are really popular. The recommended results are basically concentrated on these hundreds of products that have withstood the market test, and it is difficult to reach thousands of people. The effect of personalized configuration leads to the fact that the user's usage track data cannot reflect the user's preference well, and because the product range is limited to "explosive models", the recommended products basically cannot jump out of these hundreds of products, lack of diversity, resulting in When new insurance products come to market, there is a high probability that they will encounter the embarrassing situation of cold start

Among them, cold start refers to the situation where there is no user data precipitation in the background, and the recommendation cannot be made or the result of the recommendation is not accurate enough.

Therefore, a new product that no customers have paid attention to before will be difficult to be recommended unless special treatment is done.

Images