Principal Component Analysis (PCA) - Platform For AI - Alibaba Cloud Documentation Center

Principal component analysis (PCA) is a multivariate statistical method used to explore the internal structures of multiple variables and how they correlate to each other based on a few principal components. You can use PCA to export a few principal components that are unrelated to each other from original variables. These principal components retain as much information about the original variables as possible and are used as new comprehensive metrics.

Limits

The Principal Component Analysis (PCA) component supports only data in the dense format. You can use the component for dimensionality and noise reduction.

Configure the component

You can use one of the following methods to configure the Principal Component Analysis (PCA) component.

Method 1: Configure the component on the pipeline page

You can configure the parameters of the Principal Component Analysis (PCA) component on the pipeline page of Machine Learning Designer of Machine Learning Platform for AI (PAI). Machine Learning Designer is formerly known as Machine Learning Studio. The following table describes the parameters.


Tab	Parameter	Description
Fields Setting	Feature Columns	The columns that are selected from the input table for analysis.
Fields Setting	Appended Columns	The columns that are appended to the table after dimensionality reduction.
Parameters Setting	Data Size Ratio	The information retaining ratio after dimensionality reduction.
	Feature Decomposition Mode	The method that is used to decompose features. Valid values: CORR COVAR_SAMP COVAR_POP
	Data Conversion Method	The method that is used to convert data types. Valid values: Simple Sub-Mean Normalization
Tuning	Lifecycle	The lifecycle of the output table. The value must be a positive integer.
	Cores	The number of cores. This parameter is used with the Memory Size per Node (Unit: MB) parameter. The value must be a positive integer. Valid values: [1,9999].
	Memory Size per Node (Unit: MB)	Unit: MB. The memory size of each core. The value must be a positive integer. Valid values: [1024,64 × 1024].

Method 2: Use PAI commands

Configure the component parameters by using PAI commands. You can use the SQL Script component to call PAI commands. For more information, see SQL Script.

PAI -name PrinCompAnalysis
    -project algo_public
    -DinputTableName=bank_data
    -DeigOutputTableName=pai_temp_2032_17900_2
    -DprincompOutputTableName=pai_temp_2032_17900_1
    -DselectedColNames=pdays,previous,emp_var_rate,cons_price_idx,cons_conf_idx,euribor3m,nr_employed
    -DtransType=Simple
    -DcalcuType=CORR
    -DcontriRate=0.9;


Parameter	Required	Description	Default value
inputTableName	Yes	The input table that is used for training.	No default value
selectedColNames	Yes	The columns that are selected from the input table for analysis. Separate multiple columns with commas (,). The columns of the INT or DOUBLE data type are supported.	No default value
eigOutputTableName	Yes	The output table that contains feature vectors and feature values.	No default value
princompOutputTableName	Yes	The output table after dimensionality and noise reduction of principal components.	No default value
transType	No	The method that is used to transform the original table to a PCA table. Valid values: Simple Sub-Mean Normalization	Simple
calcuType	No	The method that is used to decompose features of the original table. Valid values: CORR COVAR_SAMP COVAR_POP	CORR
contriRate	No	The information retaining ratio after dimensionality reduction. Valid values: (0,1).	0.9
remainColumns	No	The fields that are retained from the original table after dimensionality reduction.	No default value
coreNum	No	The number of cores. This parameter is used with the memSizePerCore parameter. The value must be a positive integer. Valid values: [1,9999].	Determined by the system
memSizePerCore	No	The memory size of each core. Unit: MB. The memory size of each core. The value must be a positive integer. Valid values: [1024,64 × 1024].	Determined by the system
lifecycle	No	The lifecycle of the output table. The value must be a positive integer.	No default value

Example

Sample output tables

Data table after dimensionality reduction
Table that contains feature values and feature vectors