Mahout Random Forest RF algorithm

Randomness in random forests is reflected in: 1. Randomness of training data 2. Choosing the randomness of a split property

Can solve the problem of classification and regression, and all have good estimation performance

1. Generating a data description file

Mahout describe-p input.csv-f Input.info-d2 I 3 n i 5 n i 3 C L (description file for executing describe generated data)

2. Training model

Mahout buildforest-d input.csv-ds input.info-sl 5-p-t 5-o forest_result (generate random forest model results)

3. Testing

Mahout testforest-i input.csv-ds input.info-m forest_result-a-o predictions

-A run after the command line interface can see the analysis results, including information such as accuracy rate

-D: Data path

-DS: Data Set

-SL: Select the number of training variables each child node should have, and for the regresstion problem defaults to One-third of all variables

-NC: Optional, tree not complementary

-ms: Optional, if the branch data size is less than this value (default is 2), the tree is not detached.

-MP: Optional, if the percentage of variance of the branch data is less than this value, the tree is not detached. (in the case of a regression problem, this value is used. The default

is 1/1000 (0.001).)

-SD: Optional, seed value used to initialize the random number generator.

-P: Using partial data implementation

-T: Set how many trees are in total

-O: Output path, including decision forest model

Mahout Random Forest RF algorithm