FAQ

The <ML_HOME>/repository/conf/machine-learner.xml file includes all the ML-specific configurations. 

The default name is jdbc/WSO2ML_DB. You can change it by changing the value of the <DataSourceName> element in the <ML_HOME>/repository/conf/machine-learner.xml file. For more information, see ML-specific configurations. 

Default size is 10000. You can change it by changing the value of the <SampleSize> property within the <SummaryStatisticsSettings> element in the <ML_HOME>/repository/conf/machine-learner.xml file. For more information, see ML-specific configurations. 

By default, the <ML_HOME>/datasets/ directory holds datasets, and the default storage type is ‘file’. You can change it by changing the value of the <StorageDirectory> property within the <DatasetStorage> element in the <ML_HOME>/repository/conf/machine-learner.xml file. For more information, see ML-specific configurations. 

By default, the <ML_HOME>/models/ directory holds models, and default storage type is ‘file’. You can change it by changing the value of the <StorageDirectory> property within the <ModelStorage> element in the <ML_HOME>/repository/conf/machine-learner.xml file. For more information, see ML-specific configurations. 

WSO2 ML uses threads in a thread pool to run different tasks such as dataset summary generation and model generation etc. You can control the size of this thread pool by changing the value of the following property in the <ML_HOME>/repository/conf/machine-learner.xml file: <Property name="ml.thread.pool.size" value="100"/> For more information, see ML-specific configurations. 

You can configure WSO2 ML to send emails on the completion of a model generation. You can have a comma-separated set of email addresses as the value of the <EmailNotificationEndpoint> property in the <ML_HOME>/repository/conf/machine-learner.xml file. For more information on  configuring email support, see Enabling Email Notifications.

If you want to change the dataset storage type to HDFS, change the value of the <StorageType> property within the <DatasetStorage> element to ‘hdfs’. For more information, see ML-specific configurations. 

If you want to change the model storage type to HDFS, change the value of the <StorageType> property within the <ModelStorage> element to ‘hdfs’. For more information, see ML-specific configurations. 

If you want to store your datasets and models in an HDFS, you need to enter HDFS URL as the value of the <HdfsURL> property in the <ML_HOME>/repository/conf/machine-learner.xml file. For more information, see ML-specific configurations.  

WSO2 ML currently supports the following formats.

• CSV with comma separated values

• TSV with tab separated values

Data can be retrieved from the following sources.

• File system

• Hadoop distributed file system

• WSO2 Data Analytics Server table

It is not mandatory to have a header row. If the dataset does not have a header row, it will be indicated when you upload the dataset. Then WSO2 ML will generate a header similar to V1, V2 .. Vn. For more information see Exploring Data.

Yes, currently it is 100MB. You can change it via the following property in <ML_HOME>/bin/wso2server.bat file (for Windows) or <ML_HOME>/bin/wso2server.sh file (for Linux).

100MB = 100 x 1024 x 1024 = 104857600 Bytes

-Dorg.apache.cxf.io.CachedOutputStream.Threshold=104857600 \

Yes it does. The following algorithms are available in this version.

• Linear Regression

• Ridge Regression

• Lasso Regression

See How to Select an Algorithm in WSO2 ML for more information 

Yes it does. The classification algorithms currently supported are as follows.

• Logistic Regression with Stochastic Gradient Descent

• Logistic Regression with Limited memory Broyden-Fletcher-Goldfarb-Shanno

• Decision tree

• Random forest

• Naive bayes

See How to Select an Algorithm in WSO2 ML for more information 

At present, very primitive support is available for clustering. K-means is the only clustering algorithm supported for this ML version. We have plans to improve on this area. Please contact us if you are someone who is interested to learn more about these plans.

It supports feature selection and missing value filling/filtering.

Yes, WSO2 ML supports dataset exploration functionality with multiple visualization techniques. See Exploring Data for more information.

Once you built a model, you can view its model summary, in which you find a summary of the model evaluation. For more information, see Evaluating Models.

For classification type algorithms, generate an accuracy measurement based on the predictions made by the model for the test dataset. Test dataset is extracted from the uploaded dataset and the proportion is configurable for each analyses.

You can download a built model or publish it to WSO2 registry. For more information, see Generating Models.

WSo2 ML 1.0.0 does not support PMML format. This is in the roadmap to be provided in future versions.