Similar crowd extension algorithm based on locality sensitive Hash algorithm

Abstract

The invention provides a similar crowd extension algorithm based on a locality sensitive Hash algorithm. An open source tool datasketch is adopted to calculate original data features so as to obtain weighted minimum Hash of feature vectors of all users, the calculation amount can be greatly reduced, the calculation speed and accuracy are improved, and meanwhile the calculation cost is reduced; and in addition, a local sensitive Hash model constructed by using an open source tool datasketch can be obtained according to the memory size of the memory and the requirement of calculation accuracy, so that the accuracy of the local sensitive Hash model is high.

Description

technical field [0001] The invention relates to the technical field of information processing, in particular to an extended algorithm for similar groups of people based on a local sensitive hash algorithm. Background technique [0002] In digital marketing, how to quickly and accurately find target customer groups is a challenging task in the Internet age. Look-alike is a general term for algorithms to find the most similar groups of people starting from seed users. As an important algorithm in the field of advertising, it can help advertisers efficiently target marketing groups. [0003] The idea of ​​the common Look-alike algorithm is as follows: first, based on simple similarity calculations, such as Cosine (continuous value) or Jaccard (01 value), by calculating the similarity between two users, start from the seed user to find similar For users, this method is logically simple, but the amount of calculation is large, which makes the calculation cost high and the calcul...

AI Technical Summary

Problems solved by technology

[0003] The idea of ​​the common Look-alike algorithm is as follows: first, based on simple similarity calculations, such as Cosine (continuous value) or Jaccard (01 value), by calculating the similarity between two users, start from the seed user to find similar For users, the logic of this method is simple, but the amount of calculation is large, which makes the calculation cost high and the calculation accuracy is poor

The second method is based on logistic regression for supervised binary classification prediction. This method only requires linear calculations, and the algorithm complexity is low during online prediction. However, this method has difficulties in negative sample sampling and cold start. and other problems, resulting in poor calculation accuracy

The third is a segment-based approximate search system, that is, to tag users, aggregate user groups through tags and give candidate marketing targets. A mature tag system can bring better marketing effects, and it is very simple and easy to apply online. Fast, but this kind of system needs to spend a lot of resources to mine tags in advance. At the same time, the cost of later maintenance and improvement of the system is relatively high, and the calculation accuracy is poor.

Images