8333579a1f
Code cleanup; fixed 3 minor bugs in the settings
2018-08-31 13:10:30 -07:00
75f34853ab
Migrate to Java 1.8
2018-08-31 12:48:39 -07:00
949c8789e7
Switch HashMaps in CompetingRiskFunctions for ArrayLists
...
Provides a mild performance improvement.
2018-08-27 19:15:06 -07:00
22944115ee
Fix bug where cause specific mortality was using the cumulative hazard
...
functions instead of the CIF
2018-08-27 13:19:04 -07:00
c85cebb59f
Broke two methods in CompetingRiskErrorRateCalculator into static
...
methods in a new class.
2018-08-27 11:18:56 -07:00
e92abdab13
Add functionality to restart tree training where previously left off.
2018-08-09 16:34:10 -07:00
6d65d48844
Increased test ntree; making test more stable.
2018-08-09 16:33:57 -07:00
55eab76610
Simplied code.
2018-08-08 15:57:28 -07:00
9d9dc9ef8d
Optimize tree training so that the best split is not applied twice
2018-08-08 11:34:02 -07:00
74151b94db
Add alternative way where functions are computed only at final step.
2018-08-07 15:49:55 -07:00
d85f4eb099
Refactored competing risk combiners and differentiators into their own
...
packages.
2018-08-07 10:59:19 -07:00
bf56dfb59d
Add ability to compute different error rates.
2018-08-07 10:52:52 -07:00
d3994212b6
Fix a bug in naive mortality error measure; implement IPCW concordance
...
measure if you can provide the censoring distribution.
2018-07-26 12:45:12 -07:00
e1caef6d56
Implement naive mortality error measure
2018-07-25 15:29:09 -07:00
650579a430
Implement naive version of concordance index.
...
Note that results DO NOT MATCH with randomForestSRC; so take these
results with a grain of salt.
2018-07-25 14:18:50 -07:00
7a77851f94
WIP - Add a CompetingRiskErrorRateCalculator.
...
Note that tests fail.
2018-07-24 14:35:27 -07:00
d4853f5232
Change how trees are saved so that they are compressed.
2018-07-18 15:29:55 -07:00
dc9d20aa1a
Add ability to load gziped CSV files
2018-07-18 10:05:49 -07:00
05f9122b58
Add capability to load trees back into memory.
2018-07-17 13:54:59 -07:00
fffdfe85bf
Finish competing risk implementation. Fix a bug in tree training
...
algorithm.
2018-07-16 16:58:11 -07:00
462b0d9c35
Implement Response & GroupDifferentiators for CompetingRisk problems.
...
Also adjusted how settings are done to allow for specifying
differentiators & responses that may require arguments.
Note that CompetingRisk code is untested at this point.
2018-07-10 14:43:51 -07:00
4bbb0e0948
Fix a bug whereby FactorCovariate fails when "NA" is provided.
...
Also improved testing around this.
2018-07-06 13:33:58 -07:00
6b62ad95c3
Add support for loading datasets by CSV files.
2018-07-06 13:21:56 -07:00
fe9ff37dcf
Upgraded Settings class to allow for covariates to be built from
...
provided values.
2018-07-05 19:04:26 -07:00
b010e79269
Add basic Settings class with persistence.
2018-07-05 13:59:52 -07:00
2cdcbe6cbf
Refactor different classes into subpackages.
2018-07-05 12:59:29 -07:00
662a6cf761
Add OTFI imputation when training forest.
...
No tests have been written yet so this is still WIP.
2018-07-05 12:05:07 -07:00
Joel Therrien
c048a285a1
Merge branch '01-factors' of joel/RandomSurvivalForests into master
2018-07-04 20:27:42 +00:00
3b8952e13c
Added some tests for FactorCovariate. Moved workshop over to test
...
codebase.
2018-07-04 13:24:34 -07:00
c7298f7da6
Fix incorrect use of non-concurrent Random object in NumericCovariate.
2018-07-04 12:18:27 -07:00
e0cfed632f
Add FactorCovariate; testing required.
2018-07-04 12:18:06 -07:00
2259528c22
Small modificaton of NumericCovariate; child classes now gurantee they
...
return NumericCovariate when getParent() is called.
2018-07-04 10:54:46 -07:00
38e70dd3a1
Add BooleanCovariate
2018-07-04 10:54:07 -07:00
e96a578ac9
Refactored code to allow for a class of covariates to determine which
...
SplitRules are tested.
Most of the refactoring involved the creation of a Covariate class (one
instance per column); with SplitRule and Value being folded in as inner
classes.
2018-07-03 17:00:02 -07:00
e7af65e8fd
Fixed a bug where Splits could be generated that had an empty daughter
...
node
2018-07-03 15:15:09 -07:00
254727e594
Add support for saving trees as forest is being trained.
...
Support for loading the trees back is not yet written.
2018-07-03 12:31:08 -07:00
df35a2007a
Remove inefficient debug code previously missed.
2018-07-03 11:20:15 -07:00
5f280d09a1
Add parallel support & fix fatal bug in TreeTrainer#findBestSplitRule.
2018-07-02 23:16:20 -07:00
df7835869a
Add functionality to train a random
...
forest in serial.
2018-07-02 17:58:53 -07:00
6192643e12
Change ResponseCombiner to be a Collector that's compatible
...
with Streams.
2018-07-02 12:27:18 -07:00
3c9c78741f
Basic functinality to train a single regression tree is
...
implemented.
2018-07-01 22:22:12 -07:00
7a467207a4
Initial commit; some base classes have been defined
...
but no logic exists yet.
2018-06-29 12:04:59 -07:00
