Optimization regarding the available nodes:
- The problem connected with the models not being update as a result of source data changes has been solved.
Now PS CLEMENTINE PRO offers the function of continuous automated machine learning, which helps to effectively mitigate any model drift effects. Continuous machine learning is inspired by the biological phenomenon of natural selection. The new function is available for use in Auto Classification and Auto Prediction nodes.
![Figure 1. Model settings for the continuous automated machine learning function.](https://www.predictivesolutions.pl/uploads/artykuly/PSCP%203%203/UstawieniaModelu_EN.png)
Figure 1. Model settings for the continuous automated machine learning function.
- A new setting has been added to the GLE node on the Building Options card Use the non-negative least-squares method.
The non-negative least-squares method (NNLS) is a restricted problem of the least squares where the coefficients cannot be negative. Not all data sets are appropriate for the NNLS estimation, since it requires that there is a positive correlation or no correlation between the predictors and the forecast variable.
- A new setting has been added to the Excel source data node Rows to be scanned for column and type.
This option's numeric field means the number of the rows scanned in order to determine the column type and the data storage format. The default value is 200.
New platforms and cooperating environments
- Support for Linux RedHat version 8.3 platform.
- Cooperation with the newer R analytical software version (4.0.4).
Extended list of supported databases and other mechanisms
- Database Db2 11.5.
- Db2 Warehouse database.
- Db2 Big SQL 7.1.0 database on Cloudera Data Platform 7.1.5.
- Apache Hive 3.1.3 data warehouse on Cloudera Data Platform 7.1.5.
- Parallel processing mechanism Cloudera Impala 3.4.0 on Cloudera Data Platform 7.1.5.
- Informix 14.10 database.