Getting the Statistics

The stats() Method

The stats() method is used to display the attributes of an existing model. It accepts the {scale: "attribute"} object as an argument.

Here is how to call the stats() method:

db.predictor_name.stats({scale: "attribute"});

Where:

Name	Description
predictor_name	The name of the predictor whose statistics you want to see.
{scale: "attribute"}	The argument of the stats() method defines the type of statistics ({scale: "features"}, or {scale: "model"}, or {scale: "ensemble"}).

The stats() Method with the {scale: "features"} Parameter

Description

The db.predictor_name.stats({scale: "features"}) method is used to display the way the model encoded the data before training.

Syntax

Here is the syntax:

db.predictor_name.stats({scale: "features"});

On execution, we get:

{
  "data":[
    {
       "column" : "number_of_rooms",
       "type" : "categorical",
       "encoder" : "OneHotEncoder",
       "role" : "feature"
    }
  ]
}

Where:

Name	Description
"column"	The name of the column.
"type"	Type of the inferred data.
"encoder"	Encoder used.
"role"	Role of the column (feature or target).

Example

Let's describe the home_rentals_model model.

db.home_rentals_model.stats({scale: "features"});

On execution, we get:

{
        "data" : [
                {
                        "column" : "number_of_rooms",
                        "type" : "categorical",
                        "encoder" : "OneHotEncoder",
                        "role" : "feature"
                },
                {
                        "column" : "number_of_bathrooms",
                        "type" : "binary",
                        "encoder" : "BinaryEncoder",
                        "role" : "feature"
                },
                {
                        "column" : "sqft",
                        "type" : "float",
                        "encoder" : "NumericEncoder",
                        "role" : "feature"
                },
                {
                        "column" : "location",
                        "type" : "categorical",
                        "encoder" : "OneHotEncoder",
                        "role" : "feature"
                },
                {
                        "column" : "days_on_market",
                        "type" : "integer",
                        "encoder" : "NumericEncoder",
                        "role" : "feature"
                },
                {
                        "column" : "initial_price",
                        "type" : "integer",
                        "encoder" : "NumericEncoder",
                        "role" : "feature"
                },
                {
                        "column" : "neighborhood",
                        "type" : "categorical",
                        "encoder" : "OneHotEncoder",
                        "role" : "feature"
                },
                {
                        "column" : "rental_price",
                        "type" : "float",
                        "encoder" : "NumericEncoder",
                        "role" : "target"
                }
        ],
        "ns" : "mindsdb.home_rentals_model"
}

The stats() Method with the {scale: "model"} Parameter

Description

The db.predictor_name.stats({scale: "model"}) method is used to display the performance of the candidate models.

Syntax

Here is the syntax:

db.predictor_name.stats({scale: "model"});

On execution, we get:

{
  "data": [
    {
       "name" : "candidate_model",
       "performance" : <0.0|1.0>,
       "training_time" : <seconds>,
       "selected" : <0|1>
    }
  ]
}

Where:

Name	Description
"name"	Name of the candidate model.
"performance"	Accuracy from 0 to 1 depending on the type of the model.
"training_time"	Time elapsed for the training of the model.
"selected"	1 for the best performing model and 0 for the rest.

Example

Let's see the output for the home_rentals_model model.

db.home_rentals_model.stats({scale: "model"});

On execution, we get:

{
        "data" : [
                {
                        "name" : "Neural",
                        "performance" : 0.999,
                        "training_time" : 48.37,
                        "selected" : 0
                },
                {
                        "name" : "LightGBM",
                        "performance" : 1,
                        "training_time" : 33,
                        "selected" : 1
                },
                {
                        "name" : "Regression",
                        "performance" : 0.999,
                        "training_time" : 0.05,
                        "selected" : 0
                }
        ],
        "ns" : "mindsdb.home_rentals_model"
}

The stats() Method with the {scale: "ensemble"} Parameter

Description

The db.predictor_name.stats({scale: "ensemble"}) method is used to display the parameters used to select the best candidate model.

Syntax

Here is the syntax:

db.predictor_name.stats({scale: "ensemble"});

On execution, we get:

+-----------------+
| ensemble        |
+-----------------+
| {JSON}          |
+-----------------+

Where:

Name	Description
ensemble	Object of the JSON type describing the parameters used to select the best candidate model.

Example

Example WIP

This example is a work in progress.

Need More Info?

If you need more information on how to describe your model or understand the results, feel free to ask us on the community Slack workspace.