Insurance product dynamic pricing method based on reinforcement learning

The invention relates to the technical field of insurance information processing, in particular to an insurance product dynamic pricing method based on reinforcement learning. According to the technical scheme, the method comprises the steps of S1, grouping customer products; s2, recalling, sorting and scoring recommended products; s3, grouping customer values; s4, generating a customer group dynamic product pricing list; and S5, dynamically pricing the product. According to the invention, clients are divided into multiple groups and sub-groups; carrying out multi-path recall on recommended products of the customer subgroups, and carrying out DeepFM sorting scoring; kmeans clustering grouping is carried out according to the features such as the last insurance time of the customers from today, and multiple customer groups are clustered; counting the products with high ranking score and occurrence frequency in each customer group, and taking the products as a dynamic product pricing list of each customer group;