Applying Machine Learning for Automatic User Story Categorization in Mobile Enterprises Application Development

Matthias Heinrich Friedrich Jurisch,Stephan Böhm,Toby James-Schulz

doi:10.3991/ijim.v14i15.15263

Matthias Heinrich Friedrich Jurisch, Stephan Böhm + Show 1 more

Open Access

https://doi.org/10.3991/ijim.v14i15.15263

Copy DOI

Abstract

<p dir="ltr"><span>Mobile enterprise applications (apps) are developed in dynamic and complex environments. Hardware characteristics, operating systems and development tools are constantly changing. In larger institutions, comprehensive corporate guidelines and requirements have to be followed. In addition, larger enterprises often develop numerous apps and lack an overview of development projects. Because of the size of such companies, a comprehensive direct information exchange between developers is often not practicable. In this situation, IT support is necessary, for example to prevent unnecessary duplication of work in the development of software artifacts such as user stories, app screen designs or code features within the company. One approach to overcome these challenges is to support reusing results from previous projects by building systems to organize and analyse the knowledge base of enterprise app development projects. For such systems an automatic categorization of artifacts is required. In this work we propose using a machine learning approach to categorize user stories. The approach is evaluated on a set of user stories from real-world mobile enterprise application development projects. The results are promising and suggest that machine learning approaches can be beneficially applied to user story classification in large companies.</span></p>

Highlights

Mobile application development as a relatively new field of software development is prone to rapid changes
Our contribution is an approach that uses state-of-theart text classification methods based on machine learning, i.e., neural networks, for the classification of user stories from mobile enterprise application development
We manually categorize a set of user stories from real-world mobile enterprise application development projects

Summary

Introduction

Mobile application development as a relatively new field of software development is prone to rapid changes. Adhering to guidelines and saving time can be addressed by improving the reuse of results that already complied with enterprise requirements in previous development projects Such reuse can be supported by building platforms that automatically archive and organize software artifacts from previous development projects. A user story contains information on the actors, functions and objects relevant to a certain aspect of the system that is described This information contained in the user story is very important for building a system that can automatically make other artifacts from development projects, e.g., screen designs or code features, accessible and easy to find. Our contribution is an approach that uses state-of-theart text classification methods based on machine learning, i.e., neural networks, for the classification of user stories from mobile enterprise application development. The best F1 score achieved for this classification is 0.9, which indicates that such a machine learning classification approach can be useful in practice

Results

Discussion

Conclusion