# Metrics and time information

#### Contains

• The metric values for the training and test sets.

The table below lists the names of parameters that define the metric values to output. The values of all functions defined by these parameters are output.

Command-line version parameters Python parameters R parameters
--custom-metric custom_metric custom_loss
--loss-function --loss-function --loss-function
--eval-metric --eval-metric --eval-metric
• Information about the number of seconds of training:

• passed since the beginning
• remaining until the end

#### Format

The resulting JSON file consists of the following arrays:

#### meta

Contains basic information about the training.

Format of the array with prettified sample data:

"meta": {
"launch_mode": "Train",
"name": "second",
"iteration_count": 1000,
"learn_metrics": [
{
"name": "Precision:class=0",
"value": "Max"
},
{
"name": "Precision:class=1",
"value": "Max"
}
],
"test_sets": [
"test",
...
"testN"
],
"test_metrics": [
{
"name":"Precision:class=0",
"value":"Max"
},
{
"name":"Precision:class=1",
"value":"Max"
}
],
"learn_sets": [
"learn"
]
}

Property Type Description
launch_mode string The specified launch mode.

Possible values:
- Train — Training launch mode.
- CV — Cross-validation launch mode (for the Python cv method only).

The command-line implementation of the Cross-validation feature returns the Train value in this parameter.

name string The experiment name.

The value can be set in the --name (--name) training parameter. The default name is .
iteration_count int The maximum number of trees that can be built when solving machine learning problems.
The final number of iterations may be less than the output in this property.
learn_metrics array A list of metrics calculated for the learning dataset and information regarding the optimization method.
test_sets array The names of the arrays within the iterations array that contain the calculated values of metrics for the validation datasets.
test_metrics array A list of metrics calculated for the validation dataset and information regarding the optimization method.
name string The name of the metric.
value string The method for defining the best value of the metric. Possible values:
- Min — The smaller the value of the metric, the better.
- Max — The bigger the value of the metric, the better.
- Undefined — The best value of the metric is not defined.
- Float value — The best value of the metric is user-defined.
learn_sets array The name of the array within the iterations array that contains the calculated values of the metrics for the learning dataset.

#### iterations

Contains an array of metric values for the training and test sets and information on the duration of training for each iteration.

Format of the array with prettified sample data:

"iterations": [
{
"learn": [
0.8333333333,
0.6666666667,
0.7325581395,
-1.0836257,
0.4347826087,
0.1428571429,
0.984375,
-0.6881395691
],
"iteration": 0,
"passed_time": 0.0227411829,
"remaining_time": 22.71844172,
"test1": [
0.8333333333,
0.6666666667,
0.7325581395,
-1.0836257,
0.4347826087,
0.1428571429,
0.984375,
-0.6881395691
],
...
"testN": [
0.7333453333,
0.3666664267,
0.0325581395,
-1.9046257,
0.8937826089,
0.4138571478,
0.004313,
-0.3881390984
]
}
]

Property Type Description
learn array A list of metric values calculated for the learning dataset. The order of metrics is given in the learn_metrics array of the meta array.
iteration int The index of the iteration. Numbering starts from zero.
passed_time float The number of seconds passed since the beginning of training.
remaining_time float The number of seconds remaining until the end of training given that all the scheduled iterations take place.
test array The values of metrics calculated for the corresponding validation dataset.

The order of the metrics is given in the test_metrics array of the meta array.

#### Example

{
"meta": {
"launch_mode": "Train",
"name": "second",
"iteration_count": 1000,
"learn_metrics": [
{
"name": "Precision:class=0",
"value": "Max"
},
{
"name": "Precision:class=1",
"value": "Max"
},
{
"name": "Precision:class=2",
"value": "Max"
},
{
"name": "MultiClass",
"value": "Max"
},
{
"name": "Recall:class=0",
"value": "Max"
},
{
"name": "Recall:class=1",
"value": "Max"
},
{
"name": "Recall:class=2",
"value": "Max"
},
{
"name": "MultiClassOneVsAll",
"value": "Max"
}
],
"test_sets": [
"test"
],
"test_metrics": [
{
"name": "Precision:class=0",
"value": "Max"
},
{
"name": "Precision:class=1",
"value": "Max"
},
{
"name": "Precision:class=2",
"value": "Max"
},
{
"name": "MultiClass",
"value": "Max"
},
{
"name": "Recall:class=0",
"value": "Max"
},
{
"name": "Recall:class=1",
"value": "Max"
},
{
"name": "Recall:class=2",
"value": "Max"
},
{
"name": "MultiClassOneVsAll",
"value": "Max"
}
],
"learn_sets": [
"learn"
]
},
"iterations": [
{
"learn": [
0.8333333333,
0.6666666667,
0.7325581395,
-1.0836257,
0.4347826087,
0.1428571429,
0.984375,
-0.6881395691
],
"iteration": 0,
"passed_time": 0.0227411829,
"remaining_time": 22.71844172,
"test": [
0.8333333333,
0.6666666667,
0.7325581395,
-1.0836257,
0.4347826087,
0.1428571429,
0.984375,
-0.6881395691
]
},
{
"learn": [
0.7142857143,
1,
0.7820512821,
-1.068965402,
0.652173913,
0.1428571429,
0.953125,
-0.6832264
],
"iteration": 1,
"passed_time": 0.04471753966,
"remaining_time": 22.31405229,
"test": [
0.7142857143,
1,
0.7820512821,
-1.068965402,
0.652173913,
0.1428571429,
0.953125,
-0.6832264
]
},
...
]
}