org.bigml.binding.LocalEnsemble Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of bigml-binding Show documentation
Show all versions of bigml-binding Show documentation
An open source Java client that gives you a simple binding to interact with BigML. You can use it to
easily create, retrieve, list, update, and delete BigML resources.
/**
* A local Ensemble object.
*
* This module defines an Ensemble to make predictions locally using its
* associated models.
*
* This module can not only save you a few credits, but also enormously
* reduce the latency for each prediction and let you use your models
* offline.
*
* Example usage (assuming that you have previously set up the
* BIGML_USERNAME and BIGML_API_KEY environment variables and that you
* own the ensemble/id below):
*
*
* import org.bigml.binding.LocalEnsemble;
*
* // API client
* BigMLClient api = new BigMLClient();
*
* JSONObject ensemble = api.
* getEnsemble("ensemble/5b39e6b9c7736e583400214c");
* LocalEnsemble localEnsemble = new LocalEnsemble(ensemble)
*
* JSONObject predictors = JSONValue.parse("
* {\"petal length\": 3, \"petal width\": 0.5,
* \"sepal length\": 1, \"sepal width\": 0.5}");
*
* localEnsemble.predict(predictors)
*
*/
package org.bigml.binding;
import java.io.IOException;
import java.util.*;
import org.bigml.binding.resources.AbstractResource;
import org.bigml.binding.utils.Utils;
import org.json.simple.JSONArray;
import org.json.simple.JSONObject;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
/**
* A local predictive Ensemble.
*
* Uses a number of BigML remote models to build an ensemble local version that
* can be used to generate prediction.
*
*/
public class LocalEnsemble extends ModelFields implements SupervisedModelInterface {
private static final long serialVersionUID = 1L;
static String ENSEMBLE_RE = "^ensemble/[a-f,0-9]{24}$";
private static final int BOOSTING = 1;
private static final String[] OPERATING_POINT_KINDS = { "probability",
"confidence", "votes" };
/**
* Logging
*/
static Logger logger = LoggerFactory
.getLogger(LocalEnsemble.class.getName());
private String ensembleId;
private String objectiveField = null;
private JSONObject boosting = null;
private JSONArray models;
private JSONObject model = null;
private List modelsSplit = new ArrayList();
private String[] modelsIds;
private JSONArray distributions;
private JSONArray distribution;
private JSONObject importance;
private MultiModel multiModel;
private Boolean regression = false;
private JSONArray boostingOffsets;
private List classNames = new ArrayList();
private Map fieldNames = new HashMap();
public LocalEnsemble(JSONObject ensemble) throws Exception {
this(ensemble, null);
}
public LocalEnsemble(JSONObject ensemble, Integer maxModels)
throws Exception {
super((JSONObject) Utils.getJSONObject(ensemble, "ensemble.fields",
new JSONObject()));
// checks whether the information needed for local predictions
// is in the first argument
if (!checkModelFields(ensemble)) {
// if the fields used by the logistic regression are not
// available, use only ID to retrieve it again
ensembleId = (String) ensemble.get("resource");
boolean validId = ensembleId.matches(ENSEMBLE_RE);
if (!validId) {
throw new Exception(
ensembleId + " is not a valid resource ID.");
}
}
if (!(ensemble.containsKey("resource")
&& ensemble.get("resource") != null)) {
BigMLClient client = new BigMLClient(null, null,
BigMLClient.STORAGE);
ensemble = client.getEnsemble(ensembleId);
if ((String) ensemble.get("resource") == null) {
throw new Exception(
ensembleId + " is not a valid resource ID.");
}
}
if (ensemble.containsKey("object")
&& ensemble.get("object") instanceof JSONObject) {
ensemble = (JSONObject) ensemble.get("object");
}
ensembleId = (String) ensemble.get("resource");
if (ensemble.containsKey("ensemble")
&& ensemble.get("ensemble") instanceof JSONObject) {
JSONObject status = (JSONObject) Utils.getJSONObject(ensemble,
"status");
if (status != null && status.containsKey("code")
&& AbstractResource.FINISHED == ((Number) status
.get("code")).intValue()) {
int ensebleType = ((Long) ensemble.get("type")).intValue();
if (ensebleType == BOOSTING) {
boosting = (JSONObject) Utils.getJSONObject(ensemble,
"boosting");
}
JSONArray modelsJson = (JSONArray) ensemble.get("models");
distributions = (JSONArray) ensemble.get("distributions");
importance = (JSONObject) ensemble.get("importance");
int mn = modelsJson.size();
modelsIds = new String[mn];
for (int i = 0; i < mn; i++) {
modelsIds[i] = (String) modelsJson.get(i);
}
JSONObject fields = (JSONObject) Utils.getJSONObject(ensemble,
"ensemble.fields", new JSONObject());
objectiveField = (String) Utils.getJSONObject(ensemble,
"objective_field");
// initialize ModelFields
super.initialize((JSONObject) fields, objectiveField, null,
null, true, true, true);
} else {
throw new Exception("The lensemble isn't finished yet");
}
} else {
throw new Exception(
String.format("Cannot create the Ensemble instance. "
+ "Could not find the 'ensemble' key in "
+ "the resource:\n\n%s", ensemble));
}
init(ensemble, maxModels);
}
/**
* Constructor with a list of model references and the number of max models
* to use
*
* @param modelsIds
* the model/id of each model to be used in the ensemble
* @param maxModels
* the maximum number of models we will use in the ensemble null
* if we do not want a maxModels value
*/
public LocalEnsemble(List modelsIds, Integer maxModels) throws Exception {
this.modelsIds = (String[]) modelsIds
.toArray(new String[modelsIds.size()]);
init(null, maxModels);
}
protected void init(JSONObject ensemble, Integer maxModels)
throws Exception {
BigMLClient bigmlClient = new BigMLClient();
models = new JSONArray();
for (String id : modelsIds) {
models.add(bigmlClient.getModel(id));
}
model = (JSONObject) models.get(0);
int numberOfModels = models.size();
maxModels = maxModels != null ? maxModels : numberOfModels;
int[] items = Utils.getRange(0, numberOfModels, maxModels);
for (int item : items) {
if (item + maxModels <= numberOfModels) {
JSONArray arrayOfModels = new JSONArray();
arrayOfModels.addAll(models.subList(item, item + maxModels));
modelsSplit.add(arrayOfModels);
}
}
if (distributions != null) {
distributions = new JSONArray();
for (Object model : models) {
JSONObject treDist = (JSONObject) Utils.getJSONObject(
(JSONObject) model, "model.tree.distribution");
if (treDist != null) {
JSONObject categories = new JSONObject();
categories.put("categories", treDist);
JSONObject info = new JSONObject();
info.put("training", categories);
distributions.add(info);
} else {
distributions = new JSONArray();
break;
}
}
if (distributions.size() == 0) {
for (Object model : models) {
distributions.add((JSONObject) Utils.getJSONObject(
(JSONObject) model, "object.model.distribution"));
}
}
}
if (boosting == null) {
addModelsAttrs(model, maxModels);
}
if (fields == null) {
calculateFields();
objectiveField = (String) Utils.getJSONObject(model,
"object.objective_field");
}
if (fields != null) {
JSONObject summary = (JSONObject) Utils.getJSONObject(fields,
objectiveField + ".summary");
if (summary != null) {
if (summary.get("bins") != null) {
distribution = (JSONArray) summary.get("bins");
} else if (summary.get("counts") != null) {
distribution = (JSONArray) summary.get("counts");
} else if (summary.get("categories") != null) {
distribution = (JSONArray) summary.get("categories");
}
}
}
String optype = (String) Utils.getJSONObject(fields,
objectiveField + ".optype");
regression = "numeric".equals(optype);
if (boosting != null) {
if (regression) {
Double boostingOffset = ((Number) ensemble.get("initial_offset")).doubleValue();
boostingOffsets = new JSONArray();
boostingOffsets.add(boostingOffset);
} else {
boostingOffsets = (JSONArray) ensemble.get("initial_offsets");
}
}
if (!regression) {
JSONObject summary = (JSONObject) Utils.getJSONObject(
(JSONObject) fields.get(objectiveField), "summary");
if (summary != null) {
JSONArray categories = (JSONArray) Utils.getJSONObject(summary,
"categories", new JSONArray());
for (Object cat : categories) {
classNames.add((String) ((JSONArray) cat).get(0));
}
Collections.sort(classNames);
}
}
if (modelsSplit.size() == 1) {
multiModel = new MultiModel(models, fields, classNames);
}
}
/**
* Returns the resourceId
*/
public String getResourceId() {
return ensembleId;
}
/**
* Returns the class names
*/
public List getClassNames() {
return classNames;
}
/**
* Calculates the full list of fields used by this ensemble. It's obtained
* from the union of fields in all models of the ensemble.
*/
protected void calculateFields() {
fields = new JSONObject();
fieldNames.clear();
for (int i = 0; i < this.modelsIds.length; i++) {
JSONObject model = (JSONObject) this.models.get(i);
JSONObject fields = (JSONObject) Utils.getJSONObject(model,
"object.model.fields");
for (Object k : fields.keySet()) {
if (null != fields.get(k)) {
String fieldName = (String) ((JSONObject) fields.get(k))
.get("name");
this.fields.put(k, fields.get(k));
this.fieldNames.put((String) k, fieldName);
}
}
}
}
/**
* Adds the boosting and fields info when the ensemble is built from a list
* of models. They can be either Model objects or the model dictionary info
* structure.
*/
private void addModelsAttrs(JSONObject model, Integer maxModels) {
boolean boostedEnsemble = (Boolean) Utils.getJSONObject(model,
"object.boosted_ensemble", false);
if (boostedEnsemble) {
this.boosting = (JSONObject) Utils.getJSONObject(model,
"object.boosting", null);
}
if (this.boosting != null) {
throw new IllegalArgumentException(
"Failed to build the local ensemble. Boosted"
+ "ensembles cannot be built from a list of "
+ "boosting models.");
}
if (fields == null) {
allModelsFields(maxModels);
objectiveFieldId = (String) Utils.getJSONObject(model,
"object.objective_field", null);
}
}
/**
* Retrieves the fields used as predictors in all the ensemble models
*/
private void allModelsFields(Integer maxModels) {
try {
for (Object split : modelsSplit.get(0)) {
LocalPredictiveModel localModel = new LocalPredictiveModel(
(JSONObject) split);
fields.putAll(localModel.getFields());
}
} catch (Exception e) {}
}
/**
* Computes the predicted distributions and combines them to give the final
* predicted distribution. Depending on the method parameter probability,
* votes or the confidence are used to weight the models.
*
*/
private List combineDistributions(JSONObject inputData,
MissingStrategy missingStrategy, PredictionMethod method)
throws Exception {
if (method == null) {
method = PredictionMethod.PROBABILITY;
}
MultiVoteList votes = null;
if (modelsSplit != null && modelsSplit.size() > 1) {
// If there's more than one chunk of models, they must be
// sequentially used to generate the votes for the prediction
votes = new MultiVoteList(null);
for (Object split : modelsSplit) {
JSONArray models = (JSONArray) split;
MultiModel multiModel = new MultiModel(models, fields,
classNames);
MultiVoteList modelVotes = multiModel.generateVotesDistribution(
inputData, missingStrategy, method);
votes.extend(modelVotes);
}
} else {
// When only one group of models is found you use the
// corresponding multimodel to predict
votes = multiModel.generateVotesDistribution(inputData,
missingStrategy, method);
}
return votes.combineToDistribution(false);
}
/**
* Computes field importance based on the field importance information of
* the individual models in the ensemble.
*/
public List getFieldImportanceData() {
Map fieldImportance = new HashMap();
if (importance != null) {
fieldImportance = importance;
} else {
boolean useDistribution = false;
List importances = new ArrayList();
if (distributions != null && distributions.size() > 0) {
useDistribution = true;
for (Object item : distributions) {
JSONObject itemObj = (JSONObject) item;
useDistribution &= itemObj.containsKey("importance");
if (!useDistribution)
break;
else {
importances.add((JSONArray) itemObj.get("importance"));
}
}
}
if (useDistribution) {
for (JSONArray importance : importances) {
JSONArray importanceInfo = (JSONArray) importance;
for (Object fieldInfo : importanceInfo) {
JSONArray fieldInfoArr = (JSONArray) fieldInfo;
String fieldId = (String) fieldInfoArr.get(0);
if (!fieldImportance.containsKey(fieldId)) {
fieldImportance.put(fieldId, 0.0);
String fieldName = (String) ((JSONObject) fields
.get(fieldId)).get("name");
JSONObject fieldNameObj = new JSONObject();
fieldNameObj.put("name", fieldName);
}
fieldImportance.put(fieldId,
fieldImportance.get(fieldId)
+ ((Number) fieldInfoArr.get(1))
.doubleValue());
}
}
} else {
for (Object model : models) {
JSONObject modelObj = (JSONObject) model;
JSONArray fieldImportanceInfo = (JSONArray) Utils
.getJSONObject(modelObj, "object.model.importance");
;
for (Object fieldInfo : fieldImportanceInfo) {
JSONArray fieldInfoArr = (JSONArray) fieldInfo;
String fieldId = (String) fieldInfoArr.get(0);
if (!fieldImportance.containsKey(fieldId)) {
fieldImportance.put(fieldId, 0.0);
String fieldName = (String) ((JSONObject) fields
.get(fieldId)).get("name");
JSONObject fieldNameObj = new JSONObject();
fieldNameObj.put("name", fieldName);
}
fieldImportance.put(fieldId,
fieldImportance.get(fieldId)
+ ((Number) fieldInfoArr.get(1))
.doubleValue());
}
}
}
for (String fieldName : fieldImportance.keySet()) {
fieldImportance.put(fieldName,
fieldImportance.get(fieldName) / models.size());
}
}
List fieldImportanceOrdered = new ArrayList();
for (String fieldName : fieldImportance.keySet()) {
JSONArray fieldInfo = new JSONArray();
fieldInfo.add(fieldName);
fieldInfo.add(fieldImportance.get(fieldName));
fieldImportanceOrdered.add(fieldInfo);
}
Collections.sort(fieldImportanceOrdered, new Comparator() {
@Override
public int compare(JSONArray jsonArray, JSONArray jsonArray2) {
return (((Number) jsonArray.get(1))
.doubleValue() > ((Number) jsonArray2.get(1))
.doubleValue() ? -1 : 1);
}
});
return fieldImportanceOrdered;
}
/**
* Returns the required data distribution by adding the distributions in the
* models
*/
public JSONArray getDataDistribution(String distributionType) {
if (distributionType == null) {
distributionType = "training";
}
JSONObject categories = new JSONObject();
if (distributions != null && distributions.size() > 0) {
for (Object item : distributions) {
JSONObject modelDist = (JSONObject) item;
JSONObject summary = (JSONObject) modelDist
.get(distributionType);
JSONArray dist = new JSONArray();
if (summary != null) {
if (summary.get("bins") != null) {
dist = (JSONArray) summary.get("bins");
} else if (summary.get("counts") != null) {
dist = (JSONArray) summary.get("counts");
} else if (summary.get("categories") != null) {
dist = (JSONArray) summary.get("categories");
}
}
for (Object distr : dist) {
JSONArray distInfo = (JSONArray) distr;
String category = (String) distInfo.get(0);
Long instances = (Long) distInfo.get(1);
if (categories.containsKey(category)) {
Long current = (Long) categories.get(category);
categories.put(category, current + instances);
} else {
categories.put(category, instances);
}
}
}
}
JSONArray distribution = new JSONArray();
for (Object cat : categories.keySet()) {
String category = (String) cat;
JSONArray item = new JSONArray();
item.add(category);
item.add(categories.get(category));
distribution.add(item);
}
sortDistribution(distribution);
return distribution;
}
/**
* Prints ensemble summary. Only field importance at present.
*
*/
public String summarize() throws IOException {
StringBuilder summarize = new StringBuilder();
JSONArray distribution = getDataDistribution("training");
if (!distribution.isEmpty()) {
summarize.append("Data distribution:\n");
summarize.append(Utils.printDistribution(distribution).toString());
summarize.append("\n\n");
}
if (this.boosting == null) {
JSONArray predictions = getDataDistribution("predictions");
if (!predictions.isEmpty()) {
summarize.append("Predicted distribution:\n");
summarize.append(
Utils.printDistribution(distribution).toString());
summarize.append("\n\n");
}
}
summarize.append("Field importance:\n");
distribution = new JSONArray();
for (Object fieldItem : importance.keySet()) {
String fieldId = (String) fieldItem;
double value = (Double) importance.get(fieldId);
JSONArray item = new JSONArray();
item.add(fieldId);
item.add(value);
distribution.add(item);
}
sortDistribution(distribution);
int count = 1;
for (Object fieldItem : distribution) {
JSONArray field = (JSONArray) fieldItem;
String fieldId = (String) field.get(0);
double value = (Double) field.get(1);
summarize.append(String.format(" %s. %s: %.2f%%\n", count++,
Utils.getJSONObject(fields, fieldId + ".name"),
(Utils.roundOff(value, 4) * 100)));
}
return summarize.toString();
}
/**
* Sorting utility
*
*/
private void sortDistribution(JSONArray distribution) {
Collections.sort(distribution, new Comparator() {
@Override
public int compare(JSONArray o1, JSONArray o2) {
Object o1Val = o1.get(1);
Object o2Val = o2.get(1);
if (o1Val instanceof Number) {
o1Val = ((Number) o1Val).doubleValue();
o2Val = ((Number) o2Val).doubleValue();
}
return ((Comparable) o2Val).compareTo(o1Val);
}
});
}
/**
* Makes a prediction based on the prediction made by every model.
*
* @param inputData
* Input data to be predicted.
* @param method
* **deprecated**. Please check the operating_kind` attribute.
* Numeric key code for the following combination methods in
* classifications/regressions: 0 - majority vote (plurality)/
* average: PLURALITY_CODE 1 - confidence weighted majority vote
* / error weighted: CONFIDENCE_CODE 2 - probability weighted
* majority vote / average: PROBABILITY_CODE 3 - threshold
* filtered vote / doesn't apply: THRESHOLD_CODE
* @param options Options to be used in threshold filtered votes.
* @param missingStrategy numeric key for the individual model's prediction
* method. See the model predict method.
* @param operatingPoint
* In classification models, this is the point of the ROC curve
* where the model will be used at. The operating point can be
* defined in terms of: - the positive_class, the class that is
* important to predict accurately - its kind: probability,
* confidence or voting - its threshold: the minimum established
* for the positive_class to be predicted. The operating_point is
* then defined as a map with three attributes, e.g.:
* {"positive_class": "Iris-setosa", "kind": "probability",
* "threshold": 0.5}
* @param operatingKind
* probability", "confidence" or "votes". Sets the property that
* decides the prediction. Used only if no operating_point is
* used
* @param median
* Uses the median of each individual model's predicted node as
* individual prediction for the specified combination method.
* @param full
* Boolean that controls whether to include the prediction's
* attributes. By default, only the prediction is produced. If
* set to True, the rest of available information is added in a
* dictionary format. The dictionary keys can be: - prediction:
* the prediction value - probability: prediction's probability -
* distribution: distribution of probabilities for each of the
* objective field classes - unused_fields: list of fields in the
* input data that
*
*/
public HashMap predict(JSONObject inputData,
PredictionMethod method, Map options,
MissingStrategy missingStrategy,
JSONObject operatingPoint, String operatingKind,
Boolean median, Boolean full) throws Exception {
if (missingStrategy == null) {
missingStrategy = MissingStrategy.LAST_PREDICTION;
}
if (median == null) {
median = false;
}
if (full == null) {
full = false;
}
// Checks and cleans inputData leaving the fields used in the model
inputData = filterInputData(inputData, full);
List unusedFields = (List) inputData
.get("unusedFields");
inputData = (JSONObject) inputData.get("newInputData");
// Strips affixes for numeric values and casts to the final field type
Utils.cast(inputData, fields);
if (median && method == null) {
// predictions with median are only available with old combiners
method = PredictionMethod.PLURALITY;
}
if (method == null && operatingPoint == null && operatingKind == null
&& !median) {
// operating_point has precedence over operating_kind. If no
// combiner is set, default operating kind is "probability"
operatingKind = "probability";
}
// Operating Point
if (operatingPoint != null) {
if (regression) {
throw new IllegalArgumentException(
"The operatingPoint argument can only be"
+ " used in classifications.");
}
return predictOperating(inputData, missingStrategy, operatingPoint);
}
if (operatingKind != null) {
if (regression) {
// for regressions, operating_kind defaults to the old combiners
method = "confidence".equals(operatingKind)
? PredictionMethod.CONFIDENCE
: PredictionMethod.PLURALITY;
return predict(inputData, method, options, missingStrategy,
null, null, null, full);
} else {
// predict operating point
return predictOperatingKind(inputData, missingStrategy,
operatingKind);
}
}
MultiVote votes = null;
if (modelsSplit != null && modelsSplit.size() > 1) {
// If there's more than one chunk of models, they must be
// sequentially used to generate the votes for the prediction
votes = new MultiVote();
for (Object split : modelsSplit) {
JSONArray models = (JSONArray) split;
MultiModel multiModel = new MultiModel(models, fields, null);
MultiVote modelVotes = multiModel.generateVotes(inputData,
missingStrategy, unusedFields);
votes.extend(modelVotes);
}
} else {
// When only one group of models is found you use the
// corresponding multimodel to predict
MultiVote votesSplit = this.multiModel.generateVotes(inputData,
missingStrategy, unusedFields);
votes = new MultiVote(votesSplit.predictions, boostingOffsets);
}
if (this.boosting != null && !this.regression) {
options = new HashMap();
JSONArray categories = (JSONArray) Utils.getJSONObject(
(JSONObject) fields.get(objectiveField),
"summary.categories", new JSONArray());
options.put("categories", categories);
}
HashMap results = votes.combine(method, options);
HashMap prediction = new HashMap();
for (Object key : results.keySet()) {
prediction.put((String) key, results.get(key));
}
if (full) {
prediction.put("unused_fields", unusedFields);
}
return prediction;
}
/**
* Computes the prediction based on a user-given operating point.
*/
private HashMap predictOperating(JSONObject inputData,
MissingStrategy missingStrategy, JSONObject operatingPoint)
throws Exception {
if (missingStrategy == null) {
missingStrategy = MissingStrategy.LAST_PREDICTION;
}
Object[] operating = Utils.parseOperatingPoint(operatingPoint,
OPERATING_POINT_KINDS, classNames);
String kind = (String) operating[0];
Double threshold = (Double) operating[1];
String positiveClass = (String) operating[2];
if (!Arrays.asList(OPERATING_POINT_KINDS).contains(kind)) {
throw new IllegalArgumentException(String.format(
"Allowed operating kinds are %", OPERATING_POINT_KINDS));
}
JSONArray predictions = null;
if (kind.equals("probability")) {
predictions = predictProbability(inputData, missingStrategy);
}
if (kind.equals("confidence")) {
predictions = predictConfidence(inputData, missingStrategy);
}
if (kind.equals("votes")) {
predictions = predictVotes(inputData, missingStrategy);
}
for (Object pred : predictions) {
HashMap prediction
= (HashMap) pred;
String category = (String) prediction.get("category");
prediction.put("prediction", prediction.get("category"));
prediction.remove("category");
if (category.equals(positiveClass)
&& (Double) prediction.get(kind) > threshold) {
return prediction;
}
}
HashMap prediction
= (HashMap) predictions.get(0);
String category = (String) prediction.get("prediction");
if (category.equals(positiveClass)) {
prediction = (HashMap) predictions.get(1);
}
return prediction;
}
/**
* Computes the prediction based on a user-given operating kind, i.e,
* confidence, probability or votes.
*/
private HashMap predictOperatingKind(JSONObject inputData,
MissingStrategy missingStrategy, String operatingKind)
throws Exception {
if (missingStrategy == null) {
missingStrategy = MissingStrategy.LAST_PREDICTION;
}
String kind = operatingKind.toLowerCase();
if (boosting != null && !"probability".equals(kind)) {
throw new IllegalArgumentException(
"Only probability is allowed as operating kind "
+ "for boosted ensembles.");
}
if (!Arrays.asList(OPERATING_POINT_KINDS).contains(kind)) {
throw new IllegalArgumentException(String.format(
"Allowed operating kinds are %", OPERATING_POINT_KINDS));
}
JSONArray predictions = null;
if (kind.equals("probability")) {
predictions = predictProbability(inputData, missingStrategy);
}
if (kind.equals("confidence")) {
predictions = predictConfidence(inputData, missingStrategy);
}
if (kind.equals("votes")) {
predictions = predictVotes(inputData, missingStrategy);
}
HashMap prediction
= (HashMap) predictions.get(0);
prediction.put("prediction", prediction.get("category"));
prediction.remove("category");
return prediction;
}
/**
* For classification models, Predicts a probability for each possible
* output class, based on input values. The input fields must be a
* dictionary keyed by field name or field ID.
*
* For regressions, the output is a single element list containing the
* prediction.
*
* @param inputData
* Input data to be predicted
* @param missingStrategy
* LAST_PREDICTION|PROPORTIONAL missing strategy for missing
* fields
*/
public JSONArray predictProbability(JSONObject inputData,
MissingStrategy missingStrategy) throws Exception {
JSONArray predictions = new JSONArray();
HashMap prediction = null;
if (regression) {
prediction = predict(inputData, PredictionMethod.PROBABILITY, null,
missingStrategy, null, null, null, true);
predictions.add(prediction);
} else {
if (boosting != null) {
prediction = predict(inputData, PredictionMethod.PLURALITY,
null, missingStrategy, null, null, null, true);
JSONArray probabilities = (JSONArray) prediction
.get("probabilities");
predictions.add(probabilities);
} else {
List output = combineDistributions(inputData,
missingStrategy, null);
for (int i = 0; i < classNames.size(); i++) {
prediction = new JSONObject();
prediction.put("category", (String) classNames.get(i));
prediction.put("probability", output.get(i));
predictions.add(prediction);
}
}
}
Utils.sortPredictions(predictions, "probability", "category");
return predictions;
}
/**
* For classification models, Predicts a confidence for each possible output
* class, based on input values. The input fields must be a dictionary keyed
* by field name or field ID.
*
* For regressions, the output is a single element list containing the
* prediction.
*
* @param inputData
* Input data to be predicted
* @param missingStrategy
* LAST_PREDICTION|PROPORTIONAL missing strategy for missing
* fields
*/
private JSONArray predictConfidence(JSONObject inputData,
MissingStrategy missingStrategy) throws Exception {
if (boosting != null) {
// we use boosting probabilities as confidences also
return predictProbability(inputData, missingStrategy);
}
JSONArray predictions = new JSONArray();
HashMap prediction = null;
if (regression) {
prediction = predict(inputData, PredictionMethod.CONFIDENCE, null,
missingStrategy, null, null, null, true);
predictions.add(prediction);
} else {
List output = combineDistributions(inputData,
missingStrategy, PredictionMethod.CONFIDENCE);
for (int i = 0; i < classNames.size(); i++) {
prediction = new JSONObject();
prediction.put("category", (String) classNames.get(i));
prediction.put("confidence", output.get(i));
predictions.add(prediction);
}
}
Utils.sortPredictions(predictions, "confidence", "category");
return predictions;
}
/**
* For classification models, Predicts the votes for each possible output
* class, based on input values. The input fields must be a dictionary keyed
* by field name or field ID.
*
* For regressions, the output is a single element list containing the
* prediction.
*
* @param inputData
* Input data to be predicted
* @param missingStrategy
* LAST_PREDICTION|PROPORTIONAL missing strategy for missing
* fields
*/
private JSONArray predictVotes(JSONObject inputData,
MissingStrategy missingStrategy) throws Exception {
JSONArray predictions = new JSONArray();
HashMap prediction = null;
if (regression) {
prediction = predict(inputData, PredictionMethod.PLURALITY, null,
missingStrategy, null, null, null, true);
predictions.add(prediction);
} else {
if (boosting != null) {
throw new IllegalArgumentException(
"Votes cannot be computed for boosted ensembles.");
} else {
List output = combineDistributions(inputData,
missingStrategy, PredictionMethod.PLURALITY);
for (int i = 0; i < classNames.size(); i++) {
prediction = new JSONObject();
prediction.put("category", (String) classNames.get(i));
prediction.put("votes", output.get(i));
predictions.add(prediction);
}
}
}
Utils.sortPredictions(predictions, "votes", "category");
return predictions;
}
}
© 2015 - 2024 Weber Informatics LLC | Privacy Policy