gencfeatures
Syntax
Description
The gencfeatures function enables you to automate the feature
engineering process in the context of a machine learning workflow. Before passing tabular
training data to a classifier, you can create new features from the predictors in the data by
using gencfeatures. Use the returned data to train the
classifier.
gencfeatures allows you to generate features from variables with data
types—such as datetime, duration, and various
int types—that are not supported by most classifier training functions.
The resulting features have data types that are supported by these training functions.
To better understand the generated features, use the describe function
of the returned FeatureTransformer
object. To apply the same training set feature transformations to a test set, use the
transform function
of the FeatureTransformer object.
[
uses automated feature engineering to create Transformer,NewTbl] = gencfeatures(Tbl,ResponseVarName,q)q features from the
predictors in Tbl. The software assumes that the
ResponseVarName variable in Tbl is the response
and does not create new features from this variable. gencfeatures
returns a FeatureTransformer object (Transformer) and
a new table (NewTbl) that contains the transformed features.
By default, gencfeatures assumes that generated features are used
to train an interpretable linear model with a binary response variable. If you have a
multiclass response variable and you want to generate features to improve the accuracy of a
bagged ensemble, specify TargetLearner="bag".
[
assumes that the vector Transformer,NewTbl] = gencfeatures(Tbl,Y,q)Y is the response variable and creates new
features from the variables in Tbl.
[
uses the explanatory model Transformer,NewTbl] = gencfeatures(Tbl,formula,q)formula to determine the response variable
in Tbl and the subset of Tbl predictors from which
to create new features.
[
specifies options using one or more name-value arguments in addition to any of the input
argument combinations in previous syntaxes. For example, you can change the expected learner
type, the method for selecting new features, and the standardization method for transformed
data.Transformer,NewTbl] = gencfeatures(___,Name=Value)
Examples
Use automated feature engineering to generate new features. Train a linear classifier using the generated features. Interpret the relationship between the generated features and the trained model.
Load the patients data set. Create a table from a subset of the variables. Display the first few rows of the table.
load patients Tbl = table(Age,Diastolic,Gender,Height,SelfAssessedHealthStatus, ... Systolic,Weight,Smoker); head(Tbl)
Age Diastolic Gender Height SelfAssessedHealthStatus Systolic Weight Smoker
___ _________ __________ ______ ________________________ ________ ______ ______
38 93 {'Male' } 71 {'Excellent'} 124 176 true
43 77 {'Male' } 69 {'Fair' } 109 163 false
38 83 {'Female'} 64 {'Good' } 125 131 false
40 75 {'Female'} 67 {'Fair' } 117 133 false
49 80 {'Female'} 64 {'Good' } 122 119 false
46 70 {'Female'} 68 {'Good' } 121 142 false
33 88 {'Female'} 64 {'Good' } 130 142 true
40 82 {'Male' } 68 {'Good' } 115 180 false
Generate 10 new features from the variables in Tbl. Specify the Smoker variable as the response. By default, gencfeatures assumes that the new features will be used to train a binary linear classifier.
rng("default") % For reproducibility [T,NewTbl] = gencfeatures(Tbl,"Smoker",10)
T =
FeatureTransformer with properties:
Type: 'classification'
TargetLearner: 'linear'
NumEngineeredFeatures: 10
NumOriginalFeatures: 0
TotalNumFeatures: 10
NewTbl=100×11 table
zsc(Systolic.^2) eb8(Diastolic) q8(Systolic) eb8(Systolic) q8(Diastolic) zsc(kmd9) zsc(sin(Age)) zsc(sin(Weight)) zsc(Height-Systolic) zsc(kmc1) Smoker
________________ ______________ ____________ _____________ _____________ _________ _____________ ________________ ____________________ _________ ______
0.15379 8 6 4 8 -1.7207 0.50027 0.19202 0.40418 0.76177 true
-1.9421 2 1 1 2 -0.22056 -1.1319 -0.4009 2.3431 1.1617 false
0.30311 4 6 5 5 0.57695 0.50027 -1.037 -0.78898 -1.4456 false
-0.85785 2 2 2 2 0.83391 1.1495 1.3039 0.85162 -0.010294 false
-0.14125 3 5 4 4 1.779 -1.3083 -0.42387 -0.34154 0.99368 false
-0.28697 1 4 3 1 0.67326 1.3761 -0.72529 0.40418 1.3755 false
1.0677 6 8 6 6 -0.42521 1.5181 -0.72529 -1.5347 -1.4456 true
-1.1361 4 2 2 5 -0.79995 1.1495 -1.0225 1.2991 1.1617 false
-1.1361 3 2 2 3 -0.80136 0.46343 1.0806 1.2991 -1.208 false
-0.71693 5 3 3 6 0.37961 -0.51304 0.16741 0.55333 -1.4456 false
-1.2734 2 1 1 2 1.2572 1.3025 1.0978 1.4482 -0.010294 false
-1.1361 1 2 2 1 1.001 -1.2545 -1.2194 1.0008 -0.010294 false
0.60534 1 6 5 1 -0.98493 -0.11998 -1.211 -0.043252 -1.208 false
1.0677 8 8 6 8 -0.27307 1.4659 1.2168 -0.34154 0.24706 true
-1.2734 3 1 1 4 0.93395 -1.3633 -0.17603 1.0008 -0.010294 false
1.0677 7 8 6 8 -0.91396 -1.04 -1.2109 -0.49069 0.24706 true
⋮
T is a FeatureTransformer object that can be used to transform new data, and newTbl contains the new features generated from the Tbl data.
To better understand the generated features, use the describe object function of the FeatureTransformer object. For example, inspect the first two generated features.
describe(T,1:2)
Type IsOriginal InputVariables Transformations
___________ __________ ______________ _______________________________________________________________
zsc(Systolic.^2) Numeric false Systolic power( ,2)
Standardization with z-score (mean = 15119.54, std = 1667.5858)
eb8(Diastolic) Categorical false Diastolic Equal-width binning (number of bins = 8)
The first feature in newTbl is a numeric variable, created by first squaring the values of the Systolic variable and then converting the results to z-scores. The second feature in newTbl is a categorical variable, created by binning the values of the Diastolic variable into 8 bins of equal width.
Use the generated features to fit a linear classifier without any regularization.
Mdl = fitclinear(NewTbl,"Smoker",Lambda=0);Plot the coefficients of the predictors used to train Mdl. Note that fitclinear expands categorical predictors before fitting a model.
p = length(Mdl.Beta); [sortedCoefs,expandedIndex] = sort(Mdl.Beta,ComparisonMethod="abs"); sortedExpandedPreds = Mdl.ExpandedPredictorNames(expandedIndex); bar(sortedCoefs,Horizontal="on") yticks(1:2:p) yticklabels(sortedExpandedPreds(1:2:end)) xlabel("Coefficient") ylabel("Expanded Predictors") title("Coefficients for Expanded Predictors")
Identify the predictors whose coefficients have larger absolute values.
bigCoefs = abs(sortedCoefs) >= 4; flip(sortedExpandedPreds(bigCoefs))
ans = 1×7 cell
{'zsc(Systolic.^2)'} {'eb8(Systolic) >= 5'} {'q8(Diastolic) >= 3'} {'eb8(Diastolic) >= 3'} {'q8(Systolic) >= 6'} {'q8(Diastolic) >= 6'} {'zsc(Height-Systolic)'}
You can use partial dependence plots to analyze the categorical features whose levels have large coefficients in terms of absolute value. For example, inspect the partial dependence plot for the q8(Diastolic) variable, whose levels q8(Diastolic) >= 3 and q8(Diastolic) >= 6 have coefficients with large absolute values. These two levels correspond to noticeable changes in the predicted scores.
plotPartialDependence(Mdl,"q8(Diastolic)",Mdl.ClassNames,NewTbl);
Generate new features to improve the model accuracy for an interpretable linear model. Compare the test set accuracy of a linear model trained on the original data to the test set accuracy of a linear model trained on the transformed features.
Load the ionosphere data set. Convert the matrix of predictors X to a table.
load ionosphere
tbl = array2table(X);Partition the data into training and test sets. Use approximately 70% of the observations as training data, and 30% of the observations as test data. Partition the data using cvpartition.
rng("default") % For reproducibility of the partition cvp = cvpartition(Y,Holdout=0.3); trainIdx = training(cvp); trainTbl = tbl(training(cvp),:); trainY = Y(trainIdx); testIdx = test(cvp); testTbl = tbl(testIdx,:); testY = Y(testIdx);
Use the training data to generate 45 new features. Inspect the returned FeatureTransformer object.
[T,newTrainTbl] = gencfeatures(trainTbl,trainY,45); T
T =
FeatureTransformer with properties:
Type: 'classification'
TargetLearner: 'linear'
NumEngineeredFeatures: 45
NumOriginalFeatures: 0
TotalNumFeatures: 45
All the generated features are engineered features rather than original features in trainTbl.
Apply the transformations stored in the object T to the test data.
newTestTbl = transform(T,testTbl);
Compare the test set performances of a linear classifier trained on the original features and a linear classifier trained on the new features.
Fit a linear model without transforming the data. Check the test set performance of the model using a confusion matrix.
originalMdl = fitclinear(trainTbl,trainY); originalPredictedLabels = predict(originalMdl,testTbl); cm = confusionchart(testY,originalPredictedLabels);
confusionMatrix = cm.NormalizedValues;
originalTestAccuracy = sum(diag(confusionMatrix))/sum(confusionMatrix,"all")originalTestAccuracy = 0.8952
Fit a linear model with the transformed data. Check the test set performance of the model using a confusion matrix.
newMdl = fitclinear(newTrainTbl,trainY); newPredictedLabels = predict(newMdl,newTestTbl); newcm = confusionchart(testY,newPredictedLabels);
newConfusionMatrix = newcm.NormalizedValues;
newTestAccuracy = sum(diag(newConfusionMatrix))/sum(newConfusionMatrix,"all")newTestAccuracy = 0.9238
The linear classifier trained on the transformed data seems to outperform the linear classifier trained on the original data.
Use gencfeatures to engineer new features before training a bagged ensemble classifier. Before making predictions on new data, apply the same feature transformations to the new data set. Compare the test set performance of the ensemble that uses the engineered features to the test set performance of the ensemble that uses the original features.
Read the sample file CreditRating_Historical.dat into a table. The predictor data consists of financial ratios and industry sector information for a list of corporate customers. The response variable consists of credit ratings assigned by a rating agency. Preview the first few rows of the data set.
creditrating = readtable("CreditRating_Historical.dat");
head(creditrating) ID WC_TA RE_TA EBIT_TA MVE_BVTD S_TA Industry Rating
_____ ______ ______ _______ ________ _____ ________ _______
62394 0.013 0.104 0.036 0.447 0.142 3 {'BB' }
48608 0.232 0.335 0.062 1.969 0.281 8 {'A' }
42444 0.311 0.367 0.074 1.935 0.366 1 {'A' }
48631 0.194 0.263 0.062 1.017 0.228 4 {'BBB'}
43768 0.121 0.413 0.057 3.647 0.466 12 {'AAA'}
39255 -0.117 -0.799 0.01 0.179 0.082 4 {'CCC'}
62236 0.087 0.158 0.049 0.816 0.324 2 {'BBB'}
39354 0.005 0.181 0.034 2.597 0.388 7 {'AA' }
Because each value in the ID variable is a unique customer ID, that is, length(unique(creditrating.ID)) is equal to the number of observations in creditrating, the ID variable is a poor predictor. Remove the ID variable from the table, and convert the Industry variable to a categorical variable.
creditrating = removevars(creditrating,"ID");
creditrating.Industry = categorical(creditrating.Industry);Convert the Rating response variable to a categorical variable.
creditrating.Rating = categorical(creditrating.Rating, ... ["AAA","AA","A","BBB","BB","B","CCC"]);
Partition the data into training and test sets. Use approximately 75% of the observations as training data, and 25% of the observations as test data. Partition the data using cvpartition.
rng("default") % For reproducibility of the partition c = cvpartition(creditrating.Rating,Holdout=0.25); trainingIndices = training(c); % Indices for the training set testIndices = test(c); % Indices for the test set creditTrain = creditrating(trainingIndices,:); creditTest = creditrating(testIndices,:);
Use the training data to generate 40 new features to fit a bagged ensemble. By default, the 40 features include original features that can be used as predictors by a bagged ensemble.
[T,newCreditTrain] = gencfeatures(creditTrain,"Rating",40, ... TargetLearner="bag"); T
T =
FeatureTransformer with properties:
Type: 'classification'
TargetLearner: 'bag'
NumEngineeredFeatures: 34
NumOriginalFeatures: 6
TotalNumFeatures: 40
Create newCreditTest by applying the transformations stored in the object T to the test data.
newCreditTest = transform(T,creditTest);
Compare the test set performances of a bagged ensemble trained on the original features and a bagged ensemble trained on the new features.
Train a bagged ensemble using the original training set creditTrain. Compute the accuracy of the model on the original test set creditTest. Visualize the results using a confusion matrix.
originalMdl = fitcensemble(creditTrain,"Rating",Method="Bag"); originalTestAccuracy = 1 - loss(originalMdl,creditTest, ... "Rating",LossFun="classiferror")
originalTestAccuracy = 0.7542
predictedTestLabels = predict(originalMdl,creditTest); confusionchart(creditTest.Rating,predictedTestLabels);
Train a bagged ensemble using the transformed training set newCreditTrain. Compute the accuracy of the model on the transformed test set newCreditTest. Visualize the results using a confusion matrix.
newMdl = fitcensemble(newCreditTrain,"Rating",Method="Bag"); newTestAccuracy = 1 - loss(newMdl,newCreditTest, ... "Rating",LossFun="classiferror")
newTestAccuracy = 0.7461
newPredictedTestLabels = predict(newMdl,newCreditTest); confusionchart(newCreditTest.Rating,newPredictedTestLabels)
The bagged ensemble trained on the transformed data seems to outperform the bagged ensemble trained on the original data.
Engineer and inspect new features before training a binary support vector machine (SVM) classifier with a Gaussian kernel. Then, assess the test set performance of the classifier.
Load the ionosphere data set, which contains radar signal data. The response variable Y indicates the quality of radar returns: g indicates good quality, and b indicates bad quality. Combine the predictor and response data into one table variable.
load ionosphere
Tbl = array2table(X);
Tbl.Y = Y;
head(Tbl) X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31 X32 X33 X34 Y
__ __ _______ ________ ________ ________ ________ ________ _______ ________ _______ ________ _______ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ _____
1 0 0.99539 -0.05889 0.85243 0.02306 0.83398 -0.37708 1 0.0376 0.85243 -0.17755 0.59755 -0.44945 0.60536 -0.38223 0.84356 -0.38542 0.58212 -0.32192 0.56971 -0.29674 0.36946 -0.47357 0.56811 -0.51171 0.41078 -0.46168 0.21266 -0.3409 0.42267 -0.54487 0.18641 -0.453 {'g'}
1 0 1 -0.18829 0.93035 -0.36156 -0.10868 -0.93597 1 -0.04549 0.50874 -0.67743 0.34432 -0.69707 -0.51685 -0.97515 0.05499 -0.62237 0.33109 -1 -0.13151 -0.453 -0.18056 -0.35734 -0.20332 -0.26569 -0.20468 -0.18401 -0.1904 -0.11593 -0.16626 -0.06288 -0.13738 -0.02447 {'b'}
1 0 1 -0.03365 1 0.00485 1 -0.12062 0.88965 0.01198 0.73082 0.05346 0.85443 0.00827 0.54591 0.00299 0.83775 -0.13644 0.75535 -0.0854 0.70887 -0.27502 0.43385 -0.12062 0.57528 -0.4022 0.58984 -0.22145 0.431 -0.17365 0.60436 -0.2418 0.56045 -0.38238 {'g'}
1 0 1 -0.45161 1 1 0.71216 -1 0 0 0 0 0 0 -1 0.14516 0.54094 -0.3933 -1 -0.54467 -0.69975 1 0 0 1 0.90695 0.51613 1 1 -0.20099 0.25682 1 -0.32382 1 {'b'}
1 0 1 -0.02401 0.9414 0.06531 0.92106 -0.23255 0.77152 -0.16399 0.52798 -0.20275 0.56409 -0.00712 0.34395 -0.27457 0.5294 -0.2178 0.45107 -0.17813 0.05982 -0.35575 0.02309 -0.52879 0.03286 -0.65158 0.1329 -0.53206 0.02431 -0.62197 -0.05707 -0.59573 -0.04608 -0.65697 {'g'}
1 0 0.02337 -0.00592 -0.09924 -0.11949 -0.00763 -0.11824 0.14706 0.06637 0.03786 -0.06302 0 0 -0.04572 -0.1554 -0.00343 -0.10196 -0.11575 -0.05414 0.01838 0.03669 0.01519 0.00888 0.03513 -0.01535 -0.0324 0.09223 -0.07859 0.00732 0 0 -0.00039 0.12011 {'b'}
1 0 0.97588 -0.10602 0.94601 -0.208 0.92806 -0.2835 0.85996 -0.27342 0.79766 -0.47929 0.78225 -0.50764 0.74628 -0.61436 0.57945 -0.68086 0.37852 -0.73641 0.36324 -0.76562 0.31898 -0.79753 0.22792 -0.81634 0.13659 -0.8251 0.04606 -0.82395 -0.04262 -0.81318 -0.13832 -0.80975 {'g'}
0 0 0 0 0 0 1 -1 0 0 -1 -1 0 0 0 0 1 1 -1 -1 0 0 0 0 1 1 1 1 0 0 1 1 0 0 {'b'}
Partition the data into training and test sets. Use approximately 75% of the observations as training data, and 25% of the observations as test data. Partition the data using cvpartition.
rng("default") % For reproducibility of the partition c = cvpartition(Tbl.Y,Holdout=0.25); trainTbl = Tbl(training(c),:); testTbl = Tbl(test(c),:);
Use the training data to generate 50 features to fit a binary SVM classifier with a Gaussian kernel. By default, the 50 features include original features that can be used as predictors by an SVM classifier. Additionally, gencfeatures uses neighborhood component analysis (NCA) to reduce the set of engineered features to the most important predictors. You can use the NCA feature selection method only when the target learner is "gaussian-svm".
[Transformer,newTrainTbl] = gencfeatures(trainTbl,"Y",50, ... TargetLearner="gaussian-svm")
Transformer =
FeatureTransformer with properties:
Type: 'classification'
TargetLearner: 'gaussian-svm'
NumEngineeredFeatures: 17
NumOriginalFeatures: 33
TotalNumFeatures: 50
newTrainTbl=264×51 table
zsc(X1) zsc(X3) zsc(X4) zsc(X5) zsc(X6) zsc(X7) zsc(X8) zsc(X9) zsc(X10) zsc(X11) zsc(X12) zsc(X13) zsc(X14) zsc(X15) zsc(X16) zsc(X17) zsc(X18) zsc(X19) zsc(X20) zsc(X21) zsc(X22) zsc(X23) zsc(X24) zsc(X25) zsc(X26) zsc(X27) zsc(X28) zsc(X29) zsc(X30) zsc(X31) zsc(X32) zsc(X33) zsc(X34) zsc(X1.*X29) zsc(X10.*X21) zsc(X10.*X33) zsc(X4+X6) zsc(X5+X6) zsc(X8+X21) zsc(X1-X7) zsc(kmc2) zsc(kmc6) q13(X3) q12(X5) q15(X6) q13(X7) q15(X21) q14(X27) eb8(X5) eb8(X7) Y
_______ ________ _________ _______ ________ ________ ________ ________ ________ _________ _________ ________ ___________ ________ ________ ________ ________ ________ _________ ________ _________ _________ ________ ________ ________ _________ ________ ________ ________ ________ __________ ________ _________ ____________ _____________ _____________ __________ __________ ___________ __________ _________ _________ _______ _______ _______ _______ ________ ________ _______ _______ _____
0.35062 0.71387 -0.24103 0.48341 -0.19017 0.58078 -1.0063 0.97782 -0.26974 0.6391 -0.67348 0.31834 -1.1367 0.38853 -0.98325 0.75294 -0.77965 0.3598 -0.57266 0.38845 -0.60939 -0.012451 -0.80961 0.28949 -0.84597 -0.2517 -0.72759 -0.28242 -0.63852 0.080034 -1.0708 -0.31129 -1.046 -0.26816 0.25043 -0.038585 -0.3418 0.2362 -0.42505 -0.33933 1.0418 1.0355 12 8 7 9 9 6 7 7 {'g'}
0.35062 0.72301 -0.18451 0.77158 -0.22962 0.91433 -0.49033 0.75792 -0.32357 0.42329 -0.20389 0.73252 -0.1965 0.29679 -0.13823 0.74354 -0.27428 0.64142 -0.10244 0.6146 -0.56742 0.09598 -0.12919 0.30176 -0.62944 0.094562 -0.28922 0.096446 -0.30619 0.41 -0.47923 0.4091 -0.8911 0.14996 0.214 -0.039539 -0.3295 0.42742 0.21888 -0.65367 0.5226 0.75088 13 12 6 13 10 8 8 8 {'g'}
0.35062 0.72301 -1.1204 0.77158 1.9267 0.33603 -2.2596 -1.0149 -0.34873 -0.87362 -0.31257 -0.64512 -0.21349 -2.0888 0.17363 0.26338 -0.79564 -2.2123 -1.0155 -1.6745 1.8964 -0.63461 0.10334 1.0283 1.9591 -0.047978 1.9396 1.0838 -0.36052 -0.22117 1.9446 -1.294 2.1421 1.2396 0.19008 -0.061281 0.6801 1.8984 -3.5052 -0.10868 -1.5842 -1.337 13 12 15 7 2 7 8 7 {'b'}
0.35062 -1.2136 -0.12242 -1.375 -0.49905 -1.1101 -0.48554 -0.72188 -0.2093 -0.80643 -0.44067 -0.64512 -0.21349 -0.61619 -0.48568 -0.61727 -0.20429 -0.77473 -0.040295 -0.50751 0.034923 -0.60903 0.12046 -0.62226 0.13548 -1.1087 0.28317 -0.7878 0.053398 -0.68758 -0.0072697 -0.67106 0.21145 -0.8259 0.19351 -0.061365 -0.49849 -1.3813 -0.90111 1.2542 -1.5842 -1.337 3 1 2 2 6 2 4 4 {'b'}
0.35062 0.67519 -0.34656 0.66615 -0.69083 0.7698 -0.81803 0.69875 -0.92315 0.54191 -1.2868 0.61614 -1.2563 0.60599 -1.4924 0.32568 -1.3793 0.02881 -1.3967 0.052918 -1.5154 -0.097458 -1.4342 -0.29246 -1.4483 -0.78193 -1.3907 -0.5715 -1.5984 -0.76498 -1.5945 -0.93671 -1.8288 -0.58719 -0.089744 0.061186 -0.8284 0.032979 -0.60879 -0.51746 1.0418 1.0355 12 10 1 11 7 4 8 8 {'g'}
-2.8413 -1.2599 -0.10917 -1.1812 -0.24013 0.91433 -2.2596 -1.0149 -0.34873 -2.6482 -2.3453 -0.64512 -0.21349 -0.54564 -0.14479 1.006 2.0324 -2.2123 -1.9207 -0.53738 -0.035976 -0.63461 0.10334 1.0283 2.1431 0.88773 1.9396 -0.65143 0.038853 1.1285 1.9446 -0.67031 -0.052094 -0.6754 0.19008 -0.061281 -0.27913 -1.0579 -2.3662 -2.5471 -1.5842 -1.337 3 3 6 13 5 14 5 8 {'b'}
0.35062 0.65074 -0.27034 0.77158 -0.5507 0.91433 -0.67645 0.97782 -1.1087 0.76913 -1.1982 0.87871 -1.0489 0.84926 -1.1622 0.9786 -0.71297 0.78777 -1.2452 0.68706 -1.2068 0.53806 -1.1348 0.77353 -1.1281 0.067353 -1.1572 -0.21008 -1.2312 0.13174 -1.4278 0.078797 -1.663 -0.18832 -0.57784 -0.51689 -0.65574 0.20838 0.14089 -0.65367 1.0418 1.0355 11 12 2 13 10 7 8 8 {'g'}
0.35062 -1.2969 -0.29857 -1.1812 -0.24013 -1.0948 -0.24765 -0.78637 -0.91197 -1.684 -1.0885 -0.64512 -0.21349 -1.065 0.70198 -1.2124 0.3075 0.44948 0.507 -0.53738 -0.035976 -0.63461 0.10334 -0.93558 0.13961 -0.64684 0.073011 -0.65143 0.038853 -0.3862 0.46286 -0.82839 0.78311 -0.6754 0.19008 0.0099774 -0.42709 -1.0579 -0.73858 1.2397 -1.5842 -1.337 2 3 6 3 5 4 5 5 {'b'}
0.35062 0.72301 0.039848 0.77158 -0.63857 0.91433 -0.79731 0.97782 -1.2544 0.90098 -1.1531 0.90648 -1.2791 0.85418 -1.4394 0.83179 -1.3466 0.54942 -1.3441 0.61066 -1.5108 0.42766 -1.4494 0.27335 -1.5965 -0.1331 -1.4636 0.047912 -1.6238 -0.12466 -1.7462 -0.22743 -2.0084 0.096399 -0.66792 -0.38227 -0.48436 0.14844 -0.033402 -0.65367 1.0418 1.0355 13 12 1 13 10 7 8 8 {'g'}
0.35062 0.72301 -1.323 0.77158 -2.4069 0.91433 -2.2596 0.97782 0.41213 0.90098 -1.1484 0.96723 1.8407 0.99753 -2.3384 1.006 -0.59315 1.0392 -1.7935 1.0877 1.8964 1.0494 2.0312 1.0283 -0.64265 0.88773 -1.0301 1.0838 -1.9482 1.1285 -1.9591 1.2557 -2.2463 1.2396 1.2105 1.1115 -2.9765 -1.0579 -0.73858 -0.65367 1.0418 1.0355 13 12 1 13 15 14 8 8 {'b'}
0.35062 0.72301 0.056082 0.77158 -0.16603 0.91433 -0.35958 0.97782 -0.16462 0.90098 0.086893 0.96723 0.20409 0.99753 0.13566 1.006 0.21703 1.0392 0.60585 1.0877 0.82892 1.0494 0.90821 1.0283 0.56194 0.88773 0.78535 1.0075 1.0054 1.1285 0.62693 1.2557 0.97284 1.1554 0.437 0.22251 -0.090217 0.47081 0.79852 -0.65367 0.5226 0.75088 13 12 8 13 15 14 8 8 {'g'}
0.35062 -0.24996 -2.2139 0.77158 0.33858 -1.1655 -2.2596 0.97782 -2.4496 -0.098384 -2.3453 -0.64512 -0.21349 -2.0888 -0.89643 -1.2213 0.076204 1.0392 -1.9207 -0.53738 -0.035976 -0.63461 0.10334 -0.96039 1.9896 -0.27735 0.59845 -0.65143 0.038853 1.1285 0.44533 -0.67031 -0.052094 -0.6754 0.19008 -0.061281 -1.4561 0.81505 -2.3662 1.3064 -1.5842 -1.337 5 12 11 2 5 6 8 4 {'b'}
0.35062 0.71598 0.035661 0.77158 -0.26691 0.87035 -0.1974 0.90034 -0.30016 0.8881 -0.15385 0.79508 -0.00095871 0.90821 -0.02921 0.82498 0.22838 0.81324 0.23893 0.78923 0.19262 0.77433 0.38176 0.70892 0.43102 0.49084 0.36373 0.72129 0.34444 0.71304 0.30366 0.69599 0.21153 0.83955 0.24325 -0.0081697 -0.18761 0.40198 0.63077 -0.61223 0.5226 0.75088 12 12 4 12 11 10 8 8 {'g'}
0.35062 0.069941 -0.052561 0.11987 -0.13112 0.054366 0.1298 -0.84005 0.36726 0.2554 -0.065971 0.35614 0.079956 0.66786 0.095978 0.3326 0.37109 -0.35253 0.8869 0.33835 0.37612 0.23129 0.53951 0.1531 0.63492 -0.15329 0.56409 0.10222 0.54169 0.050904 0.49615 1.2557 0.67277 0.15634 0.70752 1.0423 -0.14691 0.0012868 0.44389 0.15675 0.5226 0.75088 6 6 8 6 8 6 7 6 {'g'}
-2.8413 0.72301 -2.3483 -1.1812 -0.24013 -1.0948 -0.24765 0.97782 1.7521 0.90098 -2.3453 -1.804 1.8407 -0.54564 -0.14479 -2.2295 2.0324 1.0392 2.0554 -2.1625 1.8964 1.0494 1.1877 -2.393 2.1431 0.88773 1.9396 1.0838 -1.9482 1.1285 1.9446 1.2557 2.1421 -0.6754 -2.6274 3.1769 -2.0283 -1.0579 -2.3662 -0.65367 0.92994 0.89071 13 3 6 3 1 14 5 5 {'b'}
-2.8413 0.72301 2.13 -1.1812 -0.24013 -1.0948 -0.24765 -3.0077 -2.4496 -0.87362 -0.31257 -0.64512 -0.21349 -2.0888 -2.3384 -2.2295 -2.0271 -2.2123 2.0554 -2.1625 1.8964 -0.63461 0.10334 -0.68235 0.16583 0.88773 -1.7099 -2.3866 2.0259 -2.5037 1.9446 -2.5963 2.1421 -0.6754 3.0075 3.1769 1.47 -1.0579 -2.3662 -0.65367 -1.5842 -1.337 13 3 6 3 1 14 5 5 {'b'}
⋮
By default, gencfeatures standardizes the original features before including them in newTrainTbl. Because it has a constant value of 0, the original X2 variable in trainTbl is not included in newTrainTbl.
unique(trainTbl.X2)
ans = 0
Inspect the first three engineered features. Note that the engineered features are stored after the 33 original features in the Transformer object. Visualize the engineered features by using a matrix of scatter plots and histograms.
featIndex = 34:36; describe(Transformer,featIndex)
Type IsOriginal InputVariables Transformations
_______ __________ ______________ ______________________________________________________________
zsc(X1.*X29) Numeric false X1, X29 X1 .* X29
Standardization with z-score (mean = 0.35269, std = 0.5222)
zsc(X10.*X21) Numeric false X10, X21 X10 .* X21
Standardization with z-score (mean = -0.067464, std = 0.35493)
zsc(X10.*X33) Numeric false X10, X33 X10 .* X33
Standardization with z-score (mean = 0.018924, std = 0.30881)
gplotmatrix(newTrainTbl{:,featIndex},[],newTrainTbl.Y,[], ...
[],[],[],"grpbars", ...
newTrainTbl.Properties.VariableNames(featIndex))
The plots can help you better understand the engineered features. For example:
The top-left plot is a histogram of the
zsc(X1.*X29)feature. This feature consists of the standardized element-wise product of the originalX1andX29features. The histogram shows that the distribution of values corresponding to good radar returns (blue) is different from the distribution of values corresponding to bad radar returns (red). For example, many of the values inzsc(X1.*X29)that correspond to bad radar returns are between –1 and –0.5.The plot in the second row, first column is a scatter plot that compares the
zsc(X1.*X29)values (along the x-axis) to thezsc(X8.*X14)values (along the y-axis). The scatter plot shows that most of thezsc(X8.*X14)values corresponding to good radar returns (blue) are greater than –1, while many of thezsc(X8.*X14)values corresponding to bad radar returns (red) are less than 1. Note that this plot contains the same information as the plot in the first row, second column, but with the axes flipped.
Create newTestTbl by applying the transformations stored in the object Transformer to the test data.
newTestTbl = transform(Transformer,testTbl);
Train an SVM classifier with a Gaussian kernel using the transformed training set newTrainTbl. Let the fitcsvm function find an appropriate scale value for the kernel function. Compute the accuracy of the model on the transformed test set newTestTbl. Visualize the results using a confusion matrix.
Mdl = fitcsvm(newTrainTbl,"Y",KernelFunction="gaussian", ... KernelScale="auto"); testAccuracy = 1 - loss(Mdl,newTestTbl,"Y", ... LossFun="classiferror")
testAccuracy = 0.9189
predictedTestLabels = predict(Mdl,newTestTbl); confusionchart(newTestTbl.Y,predictedTestLabels)
The SVM model correctly classifies most of the observations. That is, for most observations, the class predicted by the SVM model matches the true class label.
Generate features to train a linear classifier. Compute the cross-validation classification error of the model by using the crossval function.
Load the ionosphere data set, and create a table containing the predictor data.
load ionosphere
Tbl = array2table(X);Create a random partition for stratified 5-fold cross-validation.
rng("default") % For reproducibility of the partition cvp = cvpartition(Y,KFold=5);
Compute the cross-validation classification loss for a linear model trained on the original features in Tbl.
CVMdl = fitclinear(Tbl,Y,CVPartition=cvp); cvloss = kfoldLoss(CVMdl)
cvloss = 0.1339
Create the custom function myloss (shown at the end of this example). This function generates 20 features from the training data, and then applies the same training set transformations to the test data. The function then fits a linear classifier to the training data and computes the test set loss.
Note: If you use the live script file for this example, the myloss function is already included at the end of the file. Otherwise, you need to create this function at the end of your .m file or add it as a file on the MATLAB® path.
Compute the cross-validation classification loss for a linear model trained on features generated from the predictors in Tbl.
newcvloss = mean(crossval(@myloss,Tbl,Y,Partition=cvp))
newcvloss = 0.0657
function testloss = myloss(TrainTbl,trainY,TestTbl,testY) [Transformer,NewTrainTbl] = gencfeatures(TrainTbl,trainY,20); NewTestTbl = transform(Transformer,TestTbl); Mdl = fitclinear(NewTrainTbl,trainY); testloss = loss(Mdl,NewTestTbl,testY, ... LossFun="classiferror"); end
Input Arguments
Name-Value Arguments
Output Arguments
Tips
By default, when
TargetLearneris"linear"or"gaussian-svm", the software generates new features from numeric predictors by using z-scores (seeTransformedDataStandardization). You can change the type of standardization for the transformed features. However, using some method of standardization, thereby avoiding the"none"specification, is strongly recommended. Fitting linear and SVM models works best with standardized data.When you generate features to create an SVM model with good predictive performance, specify
KernelScaleas"auto"in the call tofitcsvm. This specification allows the software to find an appropriate scale value for the SVM kernel function.
