Content-based linear regression type financial product recommending method

Abstract

The invention discloses an improved financial product recommending method by a content-based linear regression mode. When a product is recommended by using the existing content-based linear method, some coefficients in the generated regression equation are negative ones and thus the characteristics are in negative correlation with product purchasing by the user,; and the negative correlation is inconsistent with the actual application scene and an overfitting phenomenon is generated. With the method disclosed by the invention, a factor causing an error is found out directly and thus improvement is carried out, so that optimized recommendation meeting the practical situation is provided.

Description

technical field [0001] The invention belongs to the technical field of big data applications, and in particular relates to a content-based linear regression wealth management product recommendation method. Background technique [0002] The financial product recommendation method is to recommend products similar to the products he liked in the past for the user based on the products he liked in the past. The recommendation method generally includes the following three steps: 1. Product feature representation: extract some features for each product to represent this product; 2. User preference learning: use the feature data of a product that a user liked (and disliked) in the past, To learn the user's preferences; 3. Recommendation generation: a process of matching user preferences and product features, recommending a set of most relevant products for this user. [0003] When using the content-based linear method for product recommendation, some coefficients appear negative i...

AI Technical Summary

Problems solved by technology

The higher the score, the better the characteristic, and it is positively correlated with public preferences. For example, in terms of flexibility and income, the score of poor flexibility is low, the score of good flexibility is high, and the score of income difference is low. A good score is high. Combined with the actual situation, the higher the numerical value of this feature, the more popular it should be. However, the coefficient of some feature k appears negative in the training results. The theoretical results show that the larger the feature score, the less likely the user is. I like this product, contrary to the actual situation, a negative number only shows that the user does not care about this characteristic index k, and cares more about other characteristic indexes, so it may appear that the product with a lower index k also has a very high score in the selection

However, a negative value will affect the recommendation accuracy and cause overfitting, because when users face products with the same eigenvalues ​​except the k index, they will most likely prefer wealth management products with high eigenvalues. An operation with a negative technical execution coefficient is incorrectly judged

Images