A method and system for similarity measurement

Abstract

The invention relates to a method and system for measuring similarity. The method for measuring the similarity comprises the following steps of a data acquiring step, to be specific acquiring elements item-a in a set a, elements item-b in a set b and the times sim(item-a, item-b) for the non-different similarity operation of elements item-a in the set a for the elements item-b in the set b; a similarity calculation step, to be specific, executing the calculation of a similarity value sim'(item-bi, item-bj) of internal elements item-bi and elements item-bj of the set b based on the formula in the specification, wherein i, j, m and n indicate mark numbers in the set, and k is a normalization factor.

Description

technical field [0001] The invention relates to the field of information processing, in particular to a method and system for similarity measurement in the field of information processing. Background technique [0002] Currently, similarity measurement is involved in many fields, such as the Internet industry, and similarity analysis is performed based on various existing similarity measurement methods. [0003] For example, in the field of personalized recommendation, etc., the server collects and stores a large amount of data on users and their operation objects, and needs to recommend relevant operation objects that users may be interested in based on the operations performed by users. The similarity between the recommended operation object and the operation object operated by the user is used to recommend the operation object with high similarity to the user. Here, in terms of similarity measurement methods, there are generally the following types. [0004] A similarit...

AI Technical Summary

Problems solved by technology

[0008] First of all, the similarity measurement method based on attribute vectors needs to know the attribute vectors of the two objects to be compared, that is, the attribute vectors need to be known. In the case of unknown attribute vectors, the similarity cannot be judged.

[0009] Second, there is the problem of low accuracy

For example, in the case of using the cosine similarity measurement method, if the attribute vectors are not independent, that is, not orthogonal, the calculated similarity is inaccurate

For example, when the attribute vector of object Aa is {x1, y1, z1}, and the attribute vector of object Bb is {x2, y2, z2}, in the case of correlation between the above attributes, that is, in the case of non-orthogonality, its calculation The accuracy of the similarity is low, and there is a lot of information loss

[0010] In addition, there is a problem of high computational complexity

The similarity measurement method based on the association relationship and the similarity measurement method based on statistics need to find the relationship between the comparison objects, and the process is relatively complicated, which makes the similarity calculation difficult.

Images