From Maisqual Private Wiki
This is the checkpoint for the meeting on the 2013.07.24 in Lille with Philippe.
- The talk we gave with Flavien at EclipseCon France 2013: Software Quality, The Eclipse Way and Beyond and the next talk I will give at EclipseCon Europe 2013..
- Talked about dimensionality reduction (see later on this page).
- Identified conferences and articles to be written.
- Start speaking about how things will go in beginning of 2014: thesis report and defence.
- Planned next actions, and end of the thesis work.
 Articles and conferences
We identified a number of conferences that could match our interests. They have been gathered in a calc document: File:Conferences.pdf
Will really try to publish at:
- ICSE 2014 (International Conference on Software Engineering), held on 01/06/2014 in Hyderabad, India. Deadline for submissions: 13/09/2013. Tracks to be followed:
- Research papers
- SEIP (Software Engineering In Practice)
- NIER (?)
- Doctoral Symposium
- MSR 2014 (Mining Software Repositories), held with ICSE 2014. No information for now.
- FASE 2014 (International Conference on Fundamental Approaches to Software Engineering), held on 05/05/2014 in Grenoble, France. Deadline for submissions: 11/10/2013.
Will try to publish at:
- JSE 2014 (International Conference on Software Engineering and Applications), held on 02/01/2014 in Zurich, Switzerland. Deadline for submissions: 05/08/2013
- ECIR 2014 (European Conference on Information Retrieval), held on 13/05/2014 in Amsterdam, Netherlands. Deadline for submissions: 14/10/2013.
- Journal: Challenges in Open Access Publishing for Software Engineering and Information Systems Fields. Deadline: 01/10/2013.
- Journal: IEEE Software Special Issue on Reflective Practice. Deadline: 20/10/2013.
These have to be matched against the following papers ideas:
- A practitionner approach to software engineering data mining
- Tools relevance:
- Shapes of Software: templates for evolution & distribution of metrics
- Dimensionality reduction (?)
 Reducting dimensions of data
On metrics clustering identify metrics that remain stable in their clusters even with different clustering methods (ward, average, single, complete...).
Selecting a single metric to represent a cluster: find the metric which is less correlated with metrics from other clusters. Practically:
- For every metric of the cluster, compute correlation with metrics from all other clusters.
- Identify the maximum correlation from this set for the given metric.
- Take the metric that has the minimum of all maxima from the given cluster's metrics.
Validate metrics set is still relevant by identifying clusters in artefacts for the full set of metrics and for the reduced set and check their stability.
Use others techniques (outliers?) and check if their results change between full and reduced set of metrics.
 Next steps
Thesis founding ends in April, 2014. Thus:
- 2014.01.15: send finalised thesis report to Philipppe.
- End of January, 2014: send finalised and reviewed report to jury.
- End of April, 2014: defend thesis.