Download Radoop Guide - RapidMiner Documentation
Transcript
RapidMiner Radoop Documentation, Release 2.3.0 Parameters: • file format (expert): The input ExampleSet will be materialized in the specified format. This setting is ignored if the input is already a table in Text or in ORC format - in this case no materalization is performed. Please note that you can force materializing in Text/ORC by using the Store operator or by setting the File Format parameter of the Radoop Nest. Materializing in ORC format requires less storage but the execution may be slower. – Default value: TEXTFILE • regression method: Various regression methods are derived by using different types of regularization: ordinary least squares or linear least squares uses no regularization; ridge regression uses L2 regularization; and Lasso uses L1 regularization. – Default value: Linear Regression • number of iterations: Number of iterations of gradient descent to run. – Default value: 100 • step size: The initial step size of SGD for the first step. Default 0.1. In subsequent steps, the step size will decrease with stepSize/sqrt(current_iteration_number). This parameter should be < 1.0. Lower step size requires higher number of iterations. In this case the algorithm will generally converge slower but results in a better model. – Default value: 0.1 • minibatch fractions: Fraction of the input data set that should be used for one iteration of SGD. Default 1.0 (corresponding to deterministic/classical gradient descent) – Default value: 1.0 • add intercept (expert): Set if the algorithm should add an intercept. – Default value: true • use feature scaling (expert): Scaling columns to unit variance as a heuristic to reduce the condition number: During the optimization process, the convergence (rate) depends on the condition number of the training dataset. Scaling the variables often reduces this condition number heuristically, thus improving the convergence rate. Without reducing the condition number, some training datasets mixing the columns with different scales may not be able to converge. Here, if useFeatureScaling is enabled, Spark will standardize the training features by dividing the variance of each column (without subtracting the mean), and train the model in the scaled space. – Default value: true • regularization parameter (expert): The regularization parameter. – Default value: 0.01 – Depends on: regression method = Linear Regression 8.8.4 Logistic Regression Synopsis This operator is a Logistic Regression Learner. It is based on the logistic regression implementation in Spark MLlib. Description Logistic regression is used to predict a binary response. Detailed information can be found in the MLlib documentation about Logistic Regression. 8.8. Modeling - Classification and Regression 52