Machine Learning with MindsDB
To get a functional MariaDB - MindsDB installation, one needs to install the following components:
- MindsDB: follow the instructions in the project's official documentation.
- Connect Storage Engine must be enabled for the integration to work. See installing the connect storage engine.
MindsDB connects to MariaDB Server via a regular user to setup a dedicated database called
mindsdb. Which user will be used is specified within MindsDB's configuration file.
For example, if MindsDB is installed locally, one can create a user called
mindsdb@localhost. MindsDB only authenticates via the mysql_native_password plugin, hence one must set a password for the user:
CREATE USER mindsdb@localhost; SET PASSWORD for mindsdb@localhost=PASSWORD("password");
The user must be granted the global FILE privilege and all privileges on the
GRANT FILE on *.* to mindsdb@localhost; GRANT ALL on mindsdb.* to mindsdb@localhost;
Assuming MindsDB is in the python path one can start up MindsDB with the following parameters:
python -m mindsdb --config=$CONFIG_PATH --api=http,mysql
$CONFIG_PATH points to the appropriate MindsDB configuration file.
Always consult the project's official documentation for up-to-date usage scenarios as MindsDB is an actively developed project.
For a step-by-step example, you can consult the following blog post.
If the connection between MindsDB and MariaDB is successful, you should see the
mindsdb database present and two tables within it:
MindsDB, as an AutoML framework does all the work when it comes to training the AI model. What is necessary is to pass it the initial data, which MindsDB retrieves via a SELECT statement. This can be done by inserting into the
INSERT INTO `predictors` (`name`, `predict`, `select_data_query`) VALUES ('bikes_model', 'count', 'SELECT * FROM test.bike_data');
The values inserted into predictors act as a command instructing MindsDB to:
- Train a model called 'bikes_model'
- From the input data, learn to predict the 'count' column.
- The input data is generated via the select statement 'SELECT * FROM test.bike_data'.
select_data_queryshould be a valid select that MindsDB can run against MariaDB.