Skip to content

Query the model from ClickHouse database

This section assumes that you have trained a new model using the ClickHouse client or MindsDB Studio. To query the model, you will need to SELECT from the model table:

SELECT
  <target_variable> AS predicted,
  <target_variable_confidence> AS confidence,
  <target_variable_explain> AS info
FROM
  mindsdb.<model_name>
WHERE
  <feature_one> AND <feature_two>

Query the model from other databases

Note that even if you have trained the model from the ClickHouse database, you will be able to query it from other databases too.

Using functions inside WHERE clause

There is an issue where ClickHouse is not parsing the functions sent inside the WHERE clause. So, for e.g toDate() or toDateTime() inside WHERE will not work. For now you can avoid using the functions until we got feedback from ClickHouse. Track the progress of this issue here.

Query example

The following example shows you how to train a new model from the ClickHouse client. The table used for training the model is the Air Pollution in Seoul timeseries dataset. MindsDB will predict the SO2 (Sulfur dioxide) value in the air based on the values added in the WHERE statement.

SELECT
   SO2 AS predicted,
   SO2_confidence AS confidence,
   SO2_explain AS info
FROM airq_predictor
WHERE (NO2 = 0.005) AND (CO = 1.2) AND (PM10 = 5)
You should get a response from MindsDB similar toas:

SO2 confidence info
0.009897379182791115 0.99 Check JSON below 
info:{
   "predicted_value": 0.009897379182791113,
   "confidence": 0.99,
   "prediction_quality": "very confident",
   "confidence_interval": [0.0059810733322441575, 0.01381368503333807], "important_missing_information": ["Address"]
}

Model predictions

Comments