SPARK2 Survival Analysis Survival regression

In Spark.ml, an accelerated failure time (AFT) model is implemented, which is a parametric survival regression model for checking data. It describes the model of the time-to-live logarithm, so it is often referred to as a logarithmic linear model of survival analysis. Unlike proportional risk models designed for the same purpose, aft models are more likely to parallelize because each instance contributes independently to the target function.

When a aftsurvivalregressionmodel is matched on a dataset with a constant non-0 column without a intercept, Spark mllib outputs 0 coefficients for a constant non-0 column. This behavior differs from R survival:: Survreg.

Importing packages

Import Org.apache.spark.sql.SparkSessionimport Org.apache.spark.sql.Datasetimport Org.apache.spark.sql.Rowimport Org.apache.spark.sql.DataFrameimport Org.apache.spark.sql.Columnimport Org.apache.spark.sql.DataFrameReaderimport Org.apache.spark.rdd.RDDimport Org.apache.spark.sql.catalyst.encoders.ExpressionEncoderimport Org.apache.spark.sql.Encoderimport Org.apache.spark.sql.DataFrameStatFunctionsimport Org.apache.spark.sql.functions._import Org.apache.spark.ml.linalg.Vectorsimport Org.apache.spark.ml.regression.AFTSurvivalRegressionimport Org.apache.spark.ml.feature.VectorAssembler

Modeling

    Val spark = Sparksession.builder (). AppName ("Spark survival regression"). config ("spark.some.config.option", "    Some-value "). Getorcreate ()//For implicit conversions as converting RDDs to dataframes import spark.implicits._       Val datalist:list[(double, double, double, double)] = List ((2, 51, 1, 1), (2, 58, 1, 1), (2, 55, 2, 1),  (2, 28, 22, 1), (1, 21, 30, 0), (1, 19, 28, 1), (2, 25, 32, 1), (2, 48, 11, 1), (2, 47, 14,  1), (2, 25, 36, 0), (2, 31, 31, 0), (1, 24, 33, 0), (1, 25, 33, 0), (2, 30, 37, 0), (2, 33,  35, 0), (1, 36, 25, 1), (1, 30, 31, 0), (1, 41, 22, 1), (2, 43, 26, 1), (2, 45, 24, 1), (2, 0), (1, +, 0), (1, +, 0), (1,,, 1), (2,,, 1), (1,, 0)) v Al data = DATALIST.TODF ("Sex", "Age", "label", "censor"). The ("label") val Colarray = Array ("Sex", "age") Val- Embler = new Vectorassembler (). SetInputcols (Colarray). Setoutputcol ("Features") Val vecdf:dataframe = assembler.transform (data) val aft = new Aftsurvi  Valregression () Val model = Aft.fit (VECDF)//Print the coefficients, intercept and scale parameter for aft survival Regression println (S "coefficients: ${model.coefficients} Intercept:" + S "${model.intercept} scale: ${model.scale } ") Val Array (coeff1, coeff2) = Model.coefficients.toArray val intercept:double = model.intercept val Scale:dou ble = model.scale val aftdf = Model.transform (VECDF)//Risk rate H (t) aftdf.selectexpr ("Sex", "Age", "label", "CENs Or "," features "," Round (prediction,2) as prediction ", S" Round (exp (sex* $coeff 1+age* $coeff $intercept), 2) as H (t) "). (" label "). Show (False)

SPARK2 Survival Analysis Survival regression