Machine Learning - Deep Learning
This method is only available in Q5.
Fits a neural network for classification or regression.
A random 30% of the data is used for cross-validation to find the optimal number of epochs according to cross-validation loss. The final network is trained on all data for the optimal number of epochs.
Usage
To run Deep Learning:
1. In Displayr, select Anything > Advanced Analysis > Machine Learning > Deep Learning. In Q, select Create > Classifier > Deep Learning.
2. Under Inputs > Deep Learning > Outcome select the outcome varaible.
3. Under Inputs > Deep Learning > Predictors select the predictor variables.
4. Change any other settings as required.
Example
The Cross Validation output of a deep learning model. Loss is the quantity minimized by the model, which is mean squared error in the case of a numeric output. mean_absolute_error is also shown.
Options
Outcome The variable to be predicted by the predictor variables. It may be either a numeric or categorical variable.
Predictors The variable(s) to predict the Outcome.
Algorithm The machine learning algorithm. Defaults to Deep Learning but may be changed to other machine learning methods.
Output
- Accuracy When Outcome is categorical, produces a table of accuracy by class. Else calculates Root Mean Squared Error and R-squared if Outcome is numeric.
- Prediction-Accuracy Table Produces a table relating the observed and predicted outcome. Also known as a confusion matrix.
- Cross Validation Produces charts of loss (i.e. network error) and accuracy or mean absolute error vs training epoch.
- Network Layers This returns a description of the layers of the network.
Missing data See Missing Data Options.
Variable names Displays Variable Names in the output.
Maximum epochs The maximum number of epochs to train the network for. The actual number of epochs may be lower if the cross-validation error stops improving.
Hidden layers A comma delimited list of the number of units in the hidden layers.
Normalize predictors Whether the predictor variables are normalized to zero mean and unit variance. This is recommended if the variables differ significantly in their ranges. Note that categorical variables are also converted to dummy variables.
Random seed Seed used to initialize the (pseudo)random number generator for the model fitting algorithm. Different seeds may lead to slightly different answers, but should normally not make a large difference.
Increase allowed output size Check this box if you encounter a warning message "The R output had size XXX MB, exceeding the 128 MB limit..." and you need to reference the output elsewhere in your document; e.g., to save predicted values to a Data Set or examine diagnostics.
Maximum allowed size for output (MB). This control only appears if Increase allowed output size is checked. Use it to set the maximum allowed size for the regression output in MegaBytes. The warning referred to above about the R output size will state the minimum size you need to increase to to return the full output. Note that having very many large outputs in one document or page may slow down the performance of your document and increase load times.
Weight. Where a weight has been set for the R Output, a new data set is generated via resampling, and this new data set is used in the estimation.
Filter The data is automatically filtered using any filters prior to estimating the model.
Additional options are available by editing the code.
DIAGNOSTICS
Prediction-Accuracy Table Creates a table showing the observed and predicted values, as a heatmap.
SAVE VARIABLE(S)
Predicted Values Creates a new variable containing predicted values for each case in the data.
Probabilities of Each Response Creates new variables containing predicted probabilities of each response.
Acknowledgements
Uses the keras package, which uses TensorFlow.
More information
See this blog post for an introduction to deep learning.
Code
var controls = [];
// ALGORITHM
var algorithm = form.comboBox({label: "Algorithm",
alternatives: ["CART", "Deep Learning", "Gradient Boosting", "Linear Discriminant Analysis",
"Random Forest", "Regression", "Support Vector Machine"],
name: "formAlgorithm", default_value: "Deep Learning",
prompt: "Machine learning or regression algorithm for fitting the model"});
controls.push(algorithm);
algorithm = algorithm.getValue();
var regressionType = "";
if (algorithm == "Regression")
{
regressionTypeControl = form.comboBox({label: "Regression type",
alternatives: ["Linear", "Binary Logit", "Ordered Logit", "Multinomial Logit", "Poisson",
"Quasi-Poisson", "NBD"],
name: "formRegressionType", default_value: "Linear",
prompt: "Select type according to outcome variable type"});
regressionType = regressionTypeControl.getValue();
controls.push(regressionTypeControl);
}
// DEFAULT CONTROLS
missing_data_options = ["Error if missing data", "Exclude cases with missing data", "Imputation (replace missing values with estimates)"];
// AMEND DEFAULT CONTROLS PER ALGORITHM
if (algorithm == "Support Vector Machine")
output_options = ["Accuracy", "Prediction-Accuracy Table", "Detail"];
if (algorithm == "Gradient Boosting")
output_options = ["Accuracy", "Importance", "Prediction-Accuracy Table", "Detail"];
if (algorithm == "Random Forest")
output_options = ["Importance", "Prediction-Accuracy Table", "Detail"];
if (algorithm == "Deep Learning")
output_options = ["Accuracy", "Prediction-Accuracy Table", "Cross Validation", "Network Layers"];
if (algorithm == "Linear Discriminant Analysis")
output_options = ["Means", "Detail", "Prediction-Accuracy Table", "Scatterplot", "Moonplot"];
if (algorithm == "CART") {
output_options = ["Sankey", "Tree", "Text", "Prediction-Accuracy Table", "Cross Validation"];
missing_data_options = ["Error if missing data", "Exclude cases with missing data",
"Use partial data", "Imputation (replace missing values with estimates)"]
}
if (algorithm == "Regression") {
if (regressionType == "Multinomial Logit")
output_options = ["Summary", "Detail", "ANOVA"];
else if (regressionType == "Linear")
output_options = ["Summary", "Detail", "ANOVA", "Relative Importance Analysis", "Shapley Regression", "Jaccard Coefficient", "Correlation", "Effects Plot"];
else
output_options = ["Summary", "Detail", "ANOVA", "Relative Importance Analysis", "Effects Plot"];
}
// COMMON CONTROLS FOR ALL ALGORITHMS
var outputControl = form.comboBox({label: "Output", prompt: "The type of output used to show the results",
alternatives: output_options, name: "formOutput",
default_value: output_options[0]});
controls.push(outputControl);
var output = outputControl.getValue();
if (algorithm == "Regression") {
if (regressionType == "Linear") {
if (output == "Jaccard Coefficient" || output == "Correlation")
missing_data_options = ["Error if missing data", "Exclude cases with missing data", "Use partial data (pairwise correlations)"];
else
missing_data_options = ["Error if missing data", "Exclude cases with missing data", "Dummy variable adjustment", "Use partial data (pairwise correlations)", "Multiple imputation"];
}
else
missing_data_options = ["Error if missing data", "Exclude cases with missing data", "Dummy variable adjustment", "Multiple imputation"];
}
var missingControl = form.comboBox({label: "Missing data",
alternatives: missing_data_options, name: "formMissing", default_value: "Exclude cases with missing data",
prompt: "Options for handling cases with missing data"});
var missing = missingControl.getValue();
controls.push(missingControl);
controls.push(form.checkBox({label: "Variable names", name: "formNames", default_value: false, prompt: "Display names instead of labels"}));
// CONTROLS FOR SPECIFIC ALGORITHMS
if (algorithm == "Support Vector Machine")
controls.push(form.textBox({label: "Cost", name: "formCost", default_value: 1, type: "number",
prompt: "High cost produces a complex model with risk of overfitting, low cost produces a simpler mode with risk of underfitting"}));
if (algorithm == "Gradient Boosting") {
controls.push(form.comboBox({label: "Booster",
alternatives: ["gbtree", "gblinear"], name: "formBooster", default_value: "gbtree",
prompt: "Boost tree or linear underlying models"}));
controls.push(form.checkBox({label: "Grid search", name: "formSearch", default_value: false,
prompt: "Search for optimal hyperparameters"}));
}
if (algorithm == "Random Forest")
if (output == "Importance")
controls.push(form.checkBox({label: "Sort by importance", name: "formImportance", default_value: true}));
if (algorithm == "Deep Learning") {
controls.push(form.numericUpDown({name:"formEpochs", label:"Maximum epochs", default_value: 10, minimum: 1, maximum: Number.MAX_SAFE_INTEGER,
prompt: "Number of rounds of training"}));
controls.push(form.textBox({name: "formHiddenLayers", label: "Hidden layers", prompt: "Comma delimited list of the number of nodes in each hidden layer", required: true}));
controls.push(form.checkBox({label: "Normalize predictors", name: "formNormalize", default_value: true,
prompt: "Normalize to zero mean and unit variance"}));
}
if (algorithm == "Linear Discriminant Analysis") {
if (output == "Scatterplot")
{
controls.push(form.colorPicker({label: "Outcome color", name: "formOutColor", default_value:"#5B9BD5"}));
controls.push(form.colorPicker({label: "Predictors color", name: "formPredColor", default_value:"#ED7D31"}));
}
controls.push(form.comboBox({label: "Prior", alternatives: ["Equal", "Observed",], name: "formPrior", default_value: "Observed",
prompt: "Probabilities of group membership"}));
}
if (algorithm == "CART") {
controls.push(form.comboBox({label: "Pruning", alternatives: ["Minimum error", "Smallest tree", "None"],
name: "formPruning", default_value: "Minimum error",
prompt: "Remove nodes after tree has been built"}));
controls.push(form.checkBox({label: "Early stopping", name: "formStopping", default_value: false,
prompt: "Stop building tree when fit does not improve"}));
controls.push(form.comboBox({label: "Predictor category labels", alternatives: ["Full labels", "Abbreviated labels", "Letters"],
name: "formPredictorCategoryLabels", default_value: "Abbreviated labels",
prompt: "Labelling of predictor categories in the tree"}));
controls.push(form.comboBox({label: "Outcome category labels", alternatives: ["Full labels", "Abbreviated labels", "Letters"],
name: "formOutcomeCategoryLabels", default_value: "Full labels",
prompt: "Labelling of outcome categories in the tree"}));
controls.push(form.checkBox({label: "Allow long-running calculations", name: "formLongRunningCalculations", default_value: false,
prompt: "Allow predictors with more than 30 categories"}));
}
var stacked_check = false;
if (algorithm == "Regression") {
if (missing == "Multiple imputation")
controls.push(form.dropBox({label: "Auxiliary variables",
types:["Variable: Numeric, Date, Money, Categorical, OrderedCategorical"],
name: "formAuxiliaryVariables", required: false, multi:true,
prompt: "Additional variables to use when imputing missing values"}));
controls.push(form.comboBox({label: "Correction", alternatives: ["None", "False Discovery Rate", "Bonferroni"], name: "formCorrection",
default_value: "None", prompt: "Multiple comparisons correction applied when computing p-values of post-hoc comparisons"}));
var is_RIA_or_shapley = output == "Relative Importance Analysis" || output == "Shapley Regression";
var is_Jaccard_or_Correlation = output == "Jaccard Coefficient" || output == "Correlation";
if (regressionType == "Linear" && missing != "Use partial data (pairwise correlations)" && missing != "Multiple imputation")
controls.push(form.checkBox({label: "Robust standard errors", name: "formRobustSE", default_value: false,
prompt: "Standard errors are robust to violations of assumption of constant variance"}));
if (is_RIA_or_shapley)
controls.push(form.checkBox({label: "Absolute importance scores", name: "formAbsoluteImportance", default_value: false,
prompt: "Show absolute instead of signed importances"}));
if (regressionType != "Multinomial Logit" && (is_RIA_or_shapley || is_Jaccard_or_Correlation || output == "Summary"))
controls.push(form.dropBox({label: "Crosstab interaction", name: "formInteraction", types:["Variable: Numeric, Date, Money, Categorical, OrderedCategorical"],
required: false, prompt: "Categorical variable to test for interaction with other variables"}));
if (regressionType !== "Multinomial Logit")
controls.push(form.numericUpDown({name : "formOutlierProportion", label:"Automated outlier removal percentage", default_value: 0,
minimum:0, maximum:49.9, increment:0.1,
prompt: "Data points removed and model refitted based on the residual values in the model using the full dataset"}));
stacked_check_box = form.checkBox({label: "Stack data", name: "formStackedData", default_value: false,
prompt: "Allow input into the Outcome control to be a single multi variable and Predictors to be a single grid variable"})
stacked_check = stacked_check_box.getValue();
controls.push(stacked_check_box);
}
controls.push(form.numericUpDown({name:"formSeed", label:"Random seed", default_value: 12321, minimum: 1, maximum: Number.MAX_SAFE_INTEGER,
prompt: "Initializes randomization for imputation and certain algorithms"}));
let allowLargeOutputsCtrl = form.checkBox({label: "Increase allowed output size",
name: "formAllowLargeOutputs", default_value: false,
prompt: "Increase the limit on the maximum size allowed for the output to fix warnings about it being too large"});
controls.push(allowLargeOutputsCtrl);
if (allowLargeOutputsCtrl.getValue())
controls.push(form.numericUpDown({name:"formMaxOutputSize", label:"Maximum allowed size for output (MB)", default_value: 128, minimum: 1, maximum: Number.MAX_SAFE_INTEGER,
prompt: "The maximum allowed size for the returned output in MB. Very large outputs may impact document performance"}));
var outcome = form.dropBox({label: "Outcome",
types: [ stacked_check ? "VariableSet: BinaryMulti, NominalMulti, OrdinalMulti, NumericMulti" : "Variable: Numeric, Date, Money, Categorical, OrderedCategorical"],
multi: false,
name: "formOutcomeVariable",
prompt: "Independent target variable to be predicted"});
var predictors = form.dropBox({label: "Predictor(s)",
types:[ stacked_check ? "VariableSet: BinaryGrid, NumericGrid" : "Variable: Numeric, Date, Money, Categorical, OrderedCategorical"],
name: "formPredictorVariables", multi: stacked_check ? false : true,
prompt: "Dependent input variables"});
controls.unshift(predictors);
controls.unshift(outcome);
form.setInputControls(controls);
var heading_text = "";
if (regressionType == "") {
heading_text = algorithm;
} else {
heading_text = regressionType + " " + algorithm;
}
if (!!form.setObjectInspectorTitle)
form.setObjectInspectorTitle(heading_text, heading_text);
else
form.setHeading(heading_text);
library(flipMultivariates)
if (get0("formAllowLargeOutputs", ifnotfound = FALSE))
QAllowLargeResultObject(1e6*get0("formMaxOutputSize"))
WarnIfVariablesSelectedFromMultipleDataSets()
model <- MachineLearning(formula = if (isTRUE(get0("formStackedData"))) as.formula(NULL) else QFormula(formOutcomeVariable ~ formPredictorVariables),
algorithm = formAlgorithm,
weights = QPopulationWeight, subset = QFilter,
missing = formMissing,
output = if (formOutput == "Shapley Regression") "Shapley regression" else formOutput,
show.labels = !formNames,
seed = get0("formSeed"),
cost = get0("formCost"),
booster = get0("formBooster"),
grid.search = get0("formSearch"),
sort.by.importance = get0("formImportance"),
hidden.nodes = get0("formHiddenLayers"),
max.epochs = get0("formEpochs"),
normalize = get0("formNormalize"),
outcome.color = get0("formOutColor"),
predictors.color = get0("formPredColor"),
prior = get0("formPrior"),
prune = get0("formPruning"),
early.stopping = get0("formStopping"),
predictor.level.treatment = get0("formPredictorCategoryLabels"),
outcome.level.treatment = get0("formOutcomeCategoryLabels"),
long.running.calculations = get0("formLongRunningCalculations"),
type = get0("formRegressionType"),
auxiliary.data = get0("formAuxiliaryVariables"),
correction = get0("formCorrection"),
robust.se = get0("formRobustSE", ifnotfound = FALSE),
importance.absolute = get0("formAbsoluteImportance"),
interaction = get0("formInteraction"),
outlier.prop.to.remove = if (get0("formRegressionType", ifnotfound = "") != "Multinomial Logit") get0("formOutlierProportion")/100 else NULL,
stacked.data.check = get0("formStackedData"),
unstacked.data = if (isTRUE(get0("formStackedData"))) list(Y = get0("formOutcomeVariable"), X = get0("formPredictorVariables")) else NULL,
use.combined.scatter = TRUE)