• Data mining/ML for Soft Eng
  • Soft Eng/Project Management
  • Software Analytics
  • Information Sources
  • Software Engineering Repositories
  • Tools
• And vice-versa? (Soft Eng for DM/ML)
  • DevOps vs MLOps
  • CRISP-DM
• Conferences and journals

Software analytics is to enable software practitioners to perform data exploration and analysis in order to obtain insightful and actionable information for data driven tasks around software and services (Zhang et al, Malets’11)

(Source: MS Research)

As with KDD (Knowledge Discovery in Databases) the idea is also to find useful information that was previously unknown!

• Version Control Systems or Soft Conf Mang, e.g., GitHub, Bitbucket
  • Artifacts: Source code, models, tests, requirements, documentation,…
  • Meta-information: What, Who, Where, When, Why
• Issue/Bug tracking systems: Bugzilla, Jira
• IDEs: Software Testing, AI assisted development
• Mailing lists (even question & answer websites)
• Google Play and AppStore e.g., reviews, starts (quality), updates…
• Simulation can be used to generate data! e.g. System Dynamics -What if? questions

(Source: I. Herraiz)

(Source:Bitergia)

(Source:Husson Univ)

• Requirements:
  • What requirements will provide most user satisfaction while maximising profit, usability, etc?
  • Effort and cost estimation with AI
• Analysis and design:
  • Can we re-organise code/modules/packages so tey are better distributed for performance/division of labour/allocation of resources, maintenance, etc?
  • Coupling and cohesion of modules

• Coding: All types of metrics can be collected from code to predict artifact defects, allocate testing effort, code clone detection, etc.
  • AI Assisted Development helping with APIs, suggesting code, etc.
  • Genetic Improvement how to plug modules or third party components automatically (e.g., video codecs)?

• Testing
  • Most testing techniques can be seen as an optimisation problem, test cases should cover (measured as LoC, branches, etc) as much as possible with the minimum no. of test cases
    • Testing automation/optimisation: generation and selection of test cases automatically (mutation testing)
    • There is also reliability estimation, when to stop testing
• Maintenance
  • Regression testing (or test we do not to break something when we fixing another)
• Phase out:
  • When is time to stop/quit maintaining a project?

• Also ML can be applied in Project Management
  • Decisions, decisions, decisions!
  • Project management is the application of knowledge, skills, tools, and techniques to project activities to meet the project requirements (PMI).
  • Multiple projects/personnel/resources
  • Iron triangle: Time, cost and scope. Pick any two!

(Source:Wikipedia)

• Decision making tasks that need to be optimised!
• And possibly multi-objective!
  • Estimation of Cost/Effort/Schedules
  • Assignment of personnel to tasks consdiring multiple projects run in parallel
• And we can help project managers with:
  • Rules, Fuzzy rules, Association Rules
  • Regression techniques
  • Visualisation of project status, early warning of problems, etc.

Also, all datasets suffer from many of the ML issues.

Pre-processing is a major step having to deal with:

• Feature selection
• Imbalance
• Data drift
• Noise
• Missing values (and imputation)
• Etc.

• Fairness
• Explainability

Open Source provides tons of data to be analysed. The most popular (Big Data!) source is GitHub

Diomidis Spinellis has maintains a list of resources for mining software engineering research https://github.com/dspinellis/awesome-msr

For project management, there is the ISBSG (International Software Benchmarking Standards Group) database but it is not open: http://isbsg.org/

• R2021 - approx. 11,600 projects and around 60 attributes related to the project management of software related projects

Small data but difficult to collect though!

OPEN repositories such as ZENODO: https://zenodo.org/

• It provides code and data with a DOI.

Ultimate Debian Database: https://udd.debian.org/

(Linus Torvalds, source: http://codequoter.myshopify.com/)

Create an empty database

$mysql -u root -p -e 'create database wekaDB;'

Run CVSAnaly2 to populate DB tables

$cvsanaly2 --db-user=root --db-password=*****  --db- database=wekaDB
Parsing log for /tmp/weka (svn)
Executing extensions

Data are ready to be used!

$ mysql -u root -p -e 'select * from wekaDB.actions limit 5;'
Enter password: 
+----+------+---------+-----------+-----------+
| id | type | file_id | commit_id | branch_id |
+----+------+---------+-----------+-----------+
|  1 | A    |       2 |         1 |         1 |
|  2 | V    |       3 |         2 |         1 |
|  3 | V    |       4 |         3 |         1 |
|  4 | V    |       5 |         4 |         1 |
|  5 | M    |       5 |         5 |         1 |
+----+------+---------+-----------+-----------+
$

And visualised:

vg-github.py --user root --passwd ***** 
      --dir /tmp/temp --removedb --ghuser USER --ghpasswd PW 
      --vgdir /home/drg/git/vizGrimoire  --isuser MetricsGrimoire

DevOps set of practices that combines SW development (Dev) and IT operations (Ops).

… to MLOps

• Machine Learning Operations (MLOps) defines language-, framework-, platform-, and infrastructure-agnostic practices to design, develop, and maintain machine learning applications. https://ml-ops.org/

• MLOps streamlines the process of taking machine learning models to production, and then maintaining and monitoring them. MLOps is a collaborative function, often comprising data scientists, devops engineers, and IT. (DataBricks)

• Machine Learning is part of many software systems and it needs to be managed

• MLflow is an open source platform to manage the ML lifecycle, including experimentation, reproducibility, deployment, and a central model registry

• https://mlflow.org/

• MLflow components:

• CRISP-DM (Cross Industry Standard Process for Data Mining)

• Predictive Model Markup Language (PMML) is a language used to represent predictive analytic models. It allows for predictive solutions to be easily shared between PMML compliant applications (Source: Wikipedia).

• ASUM-DM Analytics Solutions Unified Method for Data Mining/Predictive Analytics is a refined and extended CRISP-DM (Source: IBM)

There is a large number of related conferences (in fact, nowadays most of them seem ML conferences!:

• ICSE (Intl Conf on Soft Eng)
• ICSM (Intl Conf on Software Maintenance and Evolution)
• MSR (Mining Software Repositories)
• PROMISE (Predictive Models and Data Analytics in Software Engineering)
• EASE (Evaluation and Assessment in Software Engieering)
• Etc.

• SSBSE (Search Based Software Engineering)
• GECCO (SBSE track)
• …

Including most important SE journals:

• IEEE Trans on Soft Eng
• Springer Empirical Soft Eng
• ACM TOSEM
• Journal of Sys and Soft.
• SQJ
• IJSEKE
• Etc.