lmPredict¶

Purpose¶

Predict outcomes for test data from a linear regression model.

Format¶

predictions = lmPredict(mdl, X_test)¶

Parameters:

mdl (struct) – Instance of a lassoModel or ridgeModel structure.
X_test (NxP matrix) – The test model features, or independent variables.

Returns:

predictions (NxK matrix) – Contains one set of predictions for each lambda used for fitting the model.

Examples¶

new;
library gml;

// Specify dataset with full path
dataset = getGAUSSHome("pkgs/gml/examples/qsar_fish_toxicity.csv");

// Split data into training sets without shuffling
shuffle = "False";
{ y_train, y_test, X_train, X_test } = trainTestSplit(dataset, "LC50 ~ . ", 0.7, shuffle);

// Declare 'mdl' to be an instance of a
// lassoModel structure to hold the estimation results
struct lassoModel mdl;

// Provide lambda
lambdas = {0, 0.01, 0.05, 0.15, 0.3};

// Estimate the model with default settings
mdl = lassoFit(y_train, X_train, lambdas);

/*
** Prediction for test data
*/
y_hat = lmPredict(mdl, X_test);

// Compute and print MSE
test_mse = meanSquaredError(y_test, y_hat);

sprintf("%20s%10.3f", "test_mse", test_mse');

The above code will print the following:

===========================================================================
Model:                        Lasso     Target Variable:               LC50
Number observations:            636     Number features:                  6
===========================================================================

======================================================================
              Lambda      0.00      0.01      0.05      0.15      0.30
======================================================================

                CIC0    0.2609    0.2354    0.1331    0.0000    0.0000
              SM1_DZ    1.2362    1.2020    1.0650    0.7872    0.4546
              GATS1i   -0.6843   -0.6442   -0.4838   -0.1878    0.0000
               NdsCH    0.4999    0.4822    0.4116    0.2328    0.0000
               NdssC    0.0654    0.0616    0.0463    0.0000    0.0000
               MLOGP    0.4032    0.4091    0.4323    0.4450    0.3820
              CONST.    2.4544    2.4914    2.6394    2.8472    2.9904
======================================================================
                  DF         5         5         5         3         1
     # Non-zero Vars         6         6         6         4         2
        Training MSE     0.933     0.933     0.946     1.033     1.225
            test_mse     0.839     0.844     0.880     1.012     1.230