Done

• For last week, see this
• Spent 5 days in California for a recruitment conference, so not much done this week

Doing

• Using cross-val rather than split
• Writing a small paper for this project latex
• Creating graphs comparing other ml algorithms

Todo

• Use random forests for feature selection
• Some more prepocessing to eliminate small/irrelevant projects
• Look into temporal correlation
• Write more for the paper, to put ideas in order

Roadblocks

• Sklearn's built in cross val is giving me awful results, especially when compared to just splitting, or using weka on a subset. Trying other options now that can handle such large dataset

Admin

• Nothing for now

Done

• Started paper on building prediction: latex
• Started paper on caution with mining challenges: latex

Doing

• Writing both papers

Todo

• Some more prepocessing to eliminate small/irrelevant projects, plus feature selection
• Look into temporal correlation
• Keep writing

Roadblocks

• none

Admin

• Nothing for now

Done

• Used sklearn with random forests to predict build results for the entire dataset
• Also performed the same experiments using only the project data, i.e. no transfer learning
• For numbers, see this

Doing

• Writing a small paper for this project latex
• Creating graphs comparing other ml algorithms

Todo

• Use random forests for feature selection
• Some more prepocessing to eliminate small/irrelevant projects
• Look into temporal correlation
• Write more for the paper, to put ideas in order

Roadblocks

• I expect less progress for next week, as I'll be spending 5 days in San Francisco for a recruitment conference (o4u-e)

Admin

• Nothing for now

Done

• For last week, see this
• Predicted the outcome of a build using cros-val for 10 different algorithms, results are as follows:

For numbers go here

Table:

Doing

• Trying feature selection, so far the results hold
• Writing a small paper for this project latex

Todo

• Some more prepocessing to eliminate small/irrelevant projects
• Look into temporal correlation
• Write more for the paper, to put ideas in order

Roadblocks

• Nothing for now

Admin

• Nothing for now