Software evolution modeling method based on dynamic topic model

Abstract

The invention discloses a software evolution modeling method based on a dynamic topic model in the field of software engineering. The method includes the steps that preprocessing is performed, a time sequence document is established, a vector space model document is established, the dynamic topic model is established, a 'document-topic' matrix is converted, and a 'topic-word' matrix is converted. Developers and a project manager are helped to understand the development emphasis in all stages of time in the project development process and the evolution process of development characteristics in the whole time period, the development and maintenance time of the developers is shortened, the software developing efficiency is improved, and the method can be used for software development and maintenance.

Description

technical field [0001] The invention relates to a software modeling method, in particular to a software evolution modeling method, which belongs to the field of software development engineering. Background technique [0002] As software projects are developed, their complexity increases and so does their maintenance and understanding. The attention of developers and project managers needs to shift between different topics and tasks many times, and they often need to answer questions such as: How has this software project changed over time? Which topic or task is the focus of the current work and how much energy is spent on it? What is the proportion of various software development features (such as fixing bugs, adding functions, modifying functions, improving functions, etc.) in different time periods? What is the focus of the same development feature in different time stages? To answer such a series of questions, you need to go to the software history library (such as so...

AI Technical Summary

Problems solved by technology

However, the software evolution model it adopts has the following problems: 1) It simply uses LDA to model the documents in the entire corpus at one time, and then calculates a series of indicators through the "document-topic" matrix to characterize the evolution process of software features. Good effect, but it does not take into account the principle of "documents in a single time period are disordered, and documents between time periods are ordered" (such as the Hall Model); 2) cumbersome post-processing such as topic connection is required process (such as the Link Model)

However, the dynamic topic model assumes that the evolution of topics obeys a normal distribution, and this limitation makes it disadvantageous for large changes, that is, it is not suitable for corpora where topics change drastically.

Images