Evaluate a few methods for constructing the model
Bug #487954 reported by
Michael Wayne Goodman
This bug affects 1 person
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| Egad |
New
|
Low
|
Michael Wayne Goodman | ||
Bug Description
The Perl version of Egad had the following method for constructing a model:
+ ngrams of rule-paths were the features of a training instance
+ a string representing all the results and comparisons was the label of a training instance
+ regular expressions narrowed the range of label strings used
This method should be implemented and evaluated against at least one other method. For example:
Alternate method 1:
+ ngrams of rule-paths as features, each result as a label
+ repeat rule-paths as needed to cover all results for an item
Alternate method 2:
+ results and comparisons as features
+ rule paths as labels
+ repeat results as necessary to cover all rule paths
| Changed in egad: | |
| status: | New → Won't Fix |
| status: | Won't Fix → New |
| importance: | Undecided → Low |
| assignee: | nobody → Michael Wayne Goodman (goodmami) |
| milestone: | none → 0.2.x |
Revision history for this message
| Michael Wayne Goodman (goodmami) wrote | #1 |
To post a comment you must log in.
Alternate 2 is probably not a worthy experiment, but Alternate 1 might be. Also, different models (i.e. machine learning methods) should be tried.