Making Predictions using the ML Models¶

The `find()` Method¶

Description¶

The find() method is used to get predictions from the model table. The data is not persistent - it is returned on the fly as a result-document.

Syntax¶

Here is the syntax:

db.predictor_name.find({column: "value", column: "value"});

On execution, we get:

{
    "column_name1" : "value",
    "column_name2": "value",
    ...columns
    "select_data_query": null,
    "when_data": null,
    "<target_variable>_original": "value",
    "<target_variable>_confidence": "value",
    "<target_variable>_explain": "{\"predicted_value\": value, \"confidence\": value, \"anomaly\": null, \"truth\": null, \"confidence_lower_bound\": value \"confidence_upper_bound\": value}",
    "<target_variable>_anomaly": "value",
    "<target_variable>_min": "value",
    "<target_variable>_max": "value"
}

Where:

Expressions	Description
`"<target_variable>_original"`	The real value of the target variable from the collection.
`"<target_variable>_confidence"`	Model confidence.
`"<target_variable>_explain"`	JSON object that contains additional information, such as `predicted_value`, `confidence`, `anomaly`, `truth`, `confidence_lower_bound`, `confidence_upper_bound`.
`"<[target_variable>_anomaly"`	Model anomaly.
`"<[target_variable>_min"`	Lower bound value.
`"<[target_variable>_max"`	Upper bound value.

Example¶

Making a Single Prediction¶

The following MQL statement fetches the predicted value of the rental_price column from the home_rentals_model model. The predicted value is the rental price of a property with attributes listed as a parameter to the find() method.

db.home_rentals_model.find({sqft: "823", location: "good", neighborhood: "downtown", days_on_market: "10"});

On execution, we get:

{
    "sqft": 823,
    "location": "good",
    "neighborhood": "downtown",
    "days_on_market": 10,
    "number_of_rooms": null,
    "number_of_bathrooms": null,
    "initial_price": null,
    "rental_price": 1431.323795180614,
    "select_data_query": null,
    "when_data": null,
    "rental_price_original": null,
    "rental_price_confidence": 0.99,
    "rental_price_explain": "{\"predicted_value\": 1431.323795180614, \"confidence\": 0.99, \"anomaly\": null, \"truth\": null, \"confidence_lower_bound\": 1379.4387560440227, \"confidence_upper_bound\": 1483.2088343172054}",
    "rental_price_anomaly": null,
    "rental_price_min": 1379.4387560440227,
    "rental_price_max": 1483.2088343172054
}

Making Bulk Predictions¶

Bulk Predictions WIP

The bulk predictions is a work in progress.