Crowdsourcing software developer recommendation method based on deep learning

Abstract

The invention provides a crowdsourcing software developer recommendation method based on deep learning. The method comprises the following steps of 1, modeling tasks and developers in a crowdsourcing platform; 2, deep feature extraction based on an attention mechanism and a deep neural network; 3, task similarity calculation of a time factor is introduced; and 4, fusing the deep neural network prediction model of the task and developer features, and taking the extracted task and developer features as the input of the deep neural network prediction model. According to the scheme, the convolutional neural network is utilized to extract hidden structures and features from various types of interaction of the task and the developer to predict the score of the developer on the task, the accuracy is further improved, and the method can obtain better recommendation precision and efficiency in a large-scale crowdsourcing platform.

Description

technical field [0001] The invention relates to a method for recommending developers on a crowdsourcing platform by using a deep learning method, and belongs to the technical field of crowdsourcing software development. Background technique [0002] With the rapid development of economic globalization and Internet technology, as well as the competitive pressure from the industry, more and more software companies are turning to crowdsourcing for software development in order to reduce costs and improve efficiency. Compared with traditional software development, crowdsourcing software development uses the wisdom of groups to complete software development more efficiently and better. In the crowdsourcing software development platform, task publishers publish software development tasks with a certain amount of rewards, developers select tasks to register and submit according to their personal abilities and interests, and finally the task publisher or the platform selects the bes...

AI Technical Summary

Problems solved by technology

[0004] Traditional recommendation system methods are mainly divided into recommendation algorithms based on collaborative filtering and content-based recommendation algorithms. The recommendation algorithm based on collaborative filtering fully considers collective wisdom, and the recommendation is highly personalized and diverse, and does not depend on content information, but has The problem of data sparsity and cold start

The content-based recommendation algorithm makes up for the cold start problem from the feature level, analyzes the content itself, and establishes features, but requires strong domain knowledge, and a small number of features may not be able to accurately describe the content

In recent years, researchers have done some work in the field of crowdsourcing software developer recommendation, using a series of recommendation algorithms such as content-based methods, expertise-based methods, etc., but most of the methods are based on task or developer modeling. When there is no deep mining of features, it is difficult to better represent the features; when matching tasks and developers, machine learning algorithms are mostly used to build models, but many machine learning algorithms tend to recommend popular developers because a lot of development is involved. Due to the sparsity of data, it is difficult for the model to dig out its deeper interaction features, resulting in unsatisfactory recommendation results.

Deep learning technology has been widely used in recommendation systems. Its advantage lies in its powerful representation learning ability, which can automatically learn high-order feature interaction modes to make up for the limitations brought by artificial feature engineering. There is little research in the area of ​​developer recommendation

Some studies have proposed a software crowdsourcing developer recommendation method based on the combination of content and neural networks, which is scalable and strong, but this method only considers the characteristics of the task and ignores the characteristics of the developer, making it difficult to obtain ideal recommendation results

Images