Crowdsourcing task recommendation algorithm using outer product attention

Abstract

The invention discloses a crowdsourcing task recommendation algorithm using an outer product, which aims to recommend tasks for workers in crowdsourcing. According to the particularity of a crowdsourcing scene relative to a mainstream scene, on the basis of improving a traditional article-based collaborative filtering neural network algorithm, an attention mechanism utilizing an outer product is designed. According to the model, for timeliness of tasks and multi-role users in a crowdsourcing scene, a crowdsourcing task attribute-level feature learning mode and weighted cross entropy with strong and weak negative samples are provided respectively, and two modes of introducing a task publisher sequence are provided. The model further provides an outer product attention mechanism, and importance weights are distributed to the user sequence. According to the crowdsourcing task recommendation algorithm, an improved collaborative filtering algorithm for a crowdsourcing scene and outer product attention jointly form the crowdsourcing task recommendation algorithm, the algorithm exceeds a collaborative filtering neural network algorithm before improvement in terms of dichotomy indexes, andexceeds an attention mechanism using an inner product in terms of dichotomy indexes and recommendation list indexes, and the time consumption is within an acceptable range.

Description

[0001] 1. Technical field [0002] The present invention designs the field of recommendation algorithms, especially relates to recommending tasks for workers in crowdsourcing scenarios, specifically an item-based collaborative filtering neural network algorithm that utilizes outer product attention. [0003] 2. Background technology [0004] With the rapid development of the Internet, the amount of data is also increasing day by day, and people are increasingly helpless in the face of massive data. The recommendation system is proposed to solve the problem of information overload caused by massive data. Recommendation systems have been applied in various Internet fields in the industry. The academic community has also proposed various frameworks for recommendation algorithms, such as association rule-based recommendation, content-based recommendation, collaborative filtering, graph-based recommendation, hybrid recommendation, and so on. With the introduction of deep learning ...

AI Technical Summary

Problems solved by technology

[0010] The purpose of the present invention is to solve the problem that the existing item-based neural network collaborative filtering algorithm cannot be directly applied to the crowdsourcing scene, solve the sparsity problem and model training problem caused by timeliness by improving the feature learning part and the loss function part, A collaborative attention mechanism improves the accuracy of assigning attention weights to bisequences

[0012] Solve the sparsity problem caused by task timeliness through the attribute-level task feature learning method, and solve the training problem caused by timeliness by introducing the weighted cross-entropy loss function of strong and weak negative samples

Images