Working with PMML Execution Engine

18 Dec 201824 minutes to read

PMML Execution engine is a PMML consumer that consumes the input PMML and creates an evaluator instance for the appropriate model, which in turn invokes the respective scoring procedure for prediction when GetResult method is called. PMML Execution Engine is easy to understand and are split into separate classes for all the models within a common namespace, Syncfusion.PMML.

PMMLEvaluatorFactory

The PMMLEvaluatorFactory is a class that represents methods to instantiate respective PMML Evaluator.

Properties and Methods

Property Table

Property/Method Description
GetPMMLEvaluatorInstance(string) Returns the respective PMMLEvaluator instance for the given PMML file path
GetPMMLEvaluatorInstance(string, PMMLValidationType) Returns the PMMLEvaluator instance based on the specified PMML file path and the given PMMLValidationType
GetPMMLEvaluatorInstance(Stream) Returns the respective PMMLEvaluator instance for the given PMML file stream
GetPMMLEvaluatorInstance(Stream, PMMLValidationType) Returns the PMMLEvaluator instance based on the specified PMML file stream and the given PMMLValidationType
GetPMMLEvaluatorInstance(TextReader) Returns the respective PMMLEvaluator instance for the given PMML file as text reader object
GetPMMLEvaluatorInstance(TextReader, PMMLValidationType) Returns the PMMLEvaluator instance based on the specified PMML file as text reader object and the given PMMLValidationType
GetPMMLEvaluatorInstance(PMMLDocument) Returns the respective PMMLevaluator for the given PMMLDocument.

The following code example illustrates you the PMMLEvaluatorFactory that renders PMML file path using the GetPMMLEvaluatorInstance() method and passes it to PMMLEvaluator instance. Here a record from Iris dataset is passed as anonymous type and the predicted result is obtained.

  • C#
    • //Sample record passed as anonymous type            
var anonymousType = new           
{                
Sepal_Length = 4.3,                
Sepal_Width = 3.2,               
Petal_Length = 0.4,               
Petal_Width = 1.1            
};            
string PmmlFilePath = "../../Iris.pmml";  

//Create instance for PMML             
PMMLEvaluator PMMLEvaluator = new PMMLEvaluatorFactory().GetPMMLEvaluatorInstance(PmmlFilePath);            

//Create PMMLEvaluator instance for the specified PMML file path after validating against its schema.  
PMMLEvaluator PMMLEvaluator = new PMMLEvaluatorFactory().GetPMMLEvaluatorInstance(PmmlFilePath, PMMLValidationType.Schema);      
 
//Gets the predicted result            
PredictedResult predictedResult = PMMLEvaluator.GetResult(anonymousType, null);            
 
//Displays the predicted result            
Console.WriteLine(predictedResult.PredictedValue);


//Output

Output: Setosa
The predicted output (Setosa) obtained from PMML Execution engine gives us a clear picture that based on given sepal and petal - width and length, the iris species is more likely to be Setosa.

    NOTE

    The above code example uses PMMLEvaluator instance. In case when the model to be evaluated is known before, then you can call the model evaluator directly. PMMLValidation types are “Schema” and “None”.

    PMMLEvaluator

    The PMMLEvaluator is an abstract class and it represents the base methods and properties for all model evaluators.

    Properties and Methods

    PMMLEvaluator Public Properties/Methods

    Property/ Method Description
    PMMLDocument Gets and Sets the PMMLDocument
    PMMLModel Gets the PMMLModel
    GetResult(object, IModelOptions) Evaluates the given anonymous object as input against the scoring procedure of the PMML model and returns the PredictedResult.
    GetResult(Dictionary<string, object>, IModelOptions) Evaluates the given Dictionary object as input against the scoring procedure of the PMML model and returns the PredictedResult.
    GetResult(ExpandoObject, IModelOptions) Evaluates the given ExpandoObject as input against the scoring procedure of the PMML model and returns the PredictedResult.
    Dispose Disposes the memory occupied by objects

    Regression Model Evaluator

    Regression Model Evaluator represents the evaluator for Regression models. It inherits all the properties and methods of PMMLEvaluator.

    Regression Model

    The general purpose of regression model is to learn more about the relationship between several independent or predictor variables and a dependent variable. In the social and natural sciences multiple regression procedures are very widely used in research.

    Property Table

    Property/Method Description
    RegressionModelEvaluator(PMMLDocument) Creates Instance for Regression model evaluator.
    GetResult(object,IModelOptions) Evaluates the given input against the scoring procedure of the Regression model and returns the PredictedResult.

    Consider the Tips dataset taken from the “reshape2” R package that has information about each tip received by a waiter over a period of few months. A regression model is created based on this dataset and saved in PMML. PMML files, input dataset and a complete C# sample illustrating this model is shipped with your installer.

    Samples location:

    %LOCALAPPDATA%\Syncfusion\EssentialStudio\%version%\Common\Analytics\Regression\Tips

    The following code example illustrates the procedure to call Regression Model Evaluator directly without the help of PMMLEvaluatorFactory.

  • C#
    • //Sample data is passed as anonymous type            
var anonymousType = new            
{              
total_bill=23.68,             
sex="Male",              
smoker="No",              
day="Sun",             
time="Dinner",              
size=2            
};            
string pmmlFilePath = "../../Tips.pmml";            

//Create instance for PMML Document            
PMMLDocument pmmlDocument = new PMMLDocument(pmmlFilePath);           

//Create instance for Mining model            
RegressionModelEvaluator regressionModel = new RegressionModelEvaluator(pmmlDocument);            

//Gets the predicted result            
PredictedResult predictedResult = regressionModel.GetResult(anonymousType, null); 
regressionModel.Dispose();            

//Displays the predicted result    
Console.WriteLine(predictedResult.PredictedValue);


//Output

Output: 3.50716866881611
The predicted output (3.5) obtained from PMML Execution engine gives us a clear picture that based on given information (independent variables), the waiter may get a tip of 3.5$.In Regression Model Evaluator the term regression usually refers to the prediction of numeric values based on the Scoring procedure used in Regression Model.Here we considered the independent variables(total_bill,sex,smoker,day,time,size) and by using these independent variables we can find the result of dependent variable (tip) which is mentioned above .

    General Regression Model Evaluator

    General Regression Model Evaluator represents the evaluator for General Regression model and it inherits all the properties and methods of PMMLEvaluator.

    General Regression Model

    General Regression Model is a supervised learning model used for both classification and regression.

    Properties and Methods

    General Regression Model Public Properties or Methods

    Property/Method Description
    GeneralRegressionModelEvaluator (PMMLDocument) Creates Instance for GeneralRegression model evaluator for the given PMML document object.
    GetResult(object,IModelOptions) Evaluates the given input against the scoring procedure of the General Regression model and returns the PredictedResult

    Consider the Iris dataset taken from the “datasets” R package. A General regression model based on it is created and saved in PMML. PMML files, input dataset and a complete C# sample illustrating this model is shipped with the installer.

    Samples location:

    %LOCALAPPDATA%\Syncfusion\EssentialStudio\%version%\Common\Analytics\General Regression\Iris Logit

    The following code example illustrates the procedure to call General Regression Model Evaluator directly without the help of PMMLEvaluatorFactory.

  • C#
    • //Sample record is passed as anonymous type           
var anonymousType = new            
{                
SepalLength = 5.6,                
SepalWidth = 2.5,                
PetalLength = 3.9,                
PetalWidth = 1.1            
};            
string pmmlFilePath = "../../Iris Logit.pmml";    
        
//Create instance for PMMLDocument            
PMMLDocument pmmlDocument = new PMMLDocument(pmmlFilePath); 
           
//Create instance for GeneralRegressionModelEvaluator            
GeneralRegressionModelEvaluator generalRegression = new GeneralRegressionModelEvaluator(pmmlDocument);  
          
//Gets the predicted result            
PredictedResult predictedResult = generalRegression.GetResult(anonymousType, null);          
generalRegression.Dispose();   
  
//Displays the predicted result            
Console.WriteLine(predictedResult.PredictedValue);


//Output

Output: 1
Here the predicted output value 1 represents the Iris category “Versicolor”. (i.e.) In Iris Datset we have considered the target value as Versicolor and if the Predicted Probability value of target is above 0.5 then by Binary Classification it is predicted as 1 (Versicolor) and the value below 0.5 is predicted as 0 (Not Versicolor).

    Naive Bayes Model Evaluator

    Naive Bayes Model Evaluator represents the evaluator for the Naive Bayes model and it inherits all the properties and methods of PMMLEvaluator.

    Naive Bayes Model

    Naive Bayes classifier is a simple probabilistic classifier based on applying Bayes’ theorem with strong and naive independence assumptions between the features. Naive Bayes is a popular method for text categorization. With appropriate pre-processing, it is competitive in this domain with more advanced methods including support vector machines.

    Bayesian classifiers have been used for text classification, such as junk email filtering such as spam, author identification, topic categorization and intrusion detection or anomaly detection in computer networks.

    NaiveBayes Model Public Properties or Methods

    Property/Method Description
    NaiveBayesModelEvaluator(PMMLDocument) Creates Instance for Naive Bayes model evaluator for the given PMML document object.
    GetResult(object,IModelOptions) Evaluates the given input against the scoring procedure of the Naive Bayes model and returns the PredictedResult. Currently NaiveBayes options available to specify whether there is a need to apply Laplace smoothing or not. We can apply Laplace smoothing by setting ApplyLaplace to true.

    Consider the Iris dataset taken from the “datasets” R package. A Naive Bayes model is created here, based on it and saved in PMML. PMML files, input dataset and a complete C# sample illustrating this model is shipped with the installer.

    Samples location:

    %LOCALAPPDATA%\Syncfusion\EssentialStudio\%version%\Common\Analytics\Naive Bayes\Iris

    The following code example illustrates the procedure to call Naive Bayes model evaluator directly without the help of PMMLEvaluatorFactory.

  • C#
    • //Sample record is passed as anonymous type           
var anonymousType = new            
{                
SepalLength = 7.2,               
SepalWidth = 3.6,                
PetalLength = 6.1,               
PetalWidth = 2.5            
};            
string pmmlFilePath = "../../Iris.pmml";            
 
//Create instance for PMMLDocument            
PMMLDocument pmmlDocument = new PMMLDocument(pmmlFilePath);            
 
//Create instance for NaiveBayesModelEvaluator            
NaiveBayesModelEvaluator naiveBayes = new NaiveBayesModelEvaluator(pmmlDocument);            
 
//Gets the predicted result            
PredictedResult predictedResult = naiveBayes.GetResult(anonymousType, null);            
naiveBayes.Dispose();            
 
//Displays the predicted result            
Console.WriteLine(predictedResult.PredictedValue);


//Output

Output: Virginica
The predicted output (Virginica) obtained from PMML Execution engine gives us a clear picture that based on given sepal and petal - width and length, the iris species is more likely to be Virginica.

    Classification and Regression Tree Model Evaluator

    Tree Model Evaluator represents the evaluator for tree models and it inherits all the properties and methods of PMMLEvaluator.

    Tree Model Public Properties or Methods

    Property/Method Description
    TreeModelEvaluator(PMMLDocument) Creates Instance for Tree model evaluator for the given PMML document object.
    GetResult(object,IModelOptions) Evaluates the given input against the scoring procedure of the Tree model and returns the PredictedResultCurrently there is no Model options available for Tree Model.

    Consider the Iris dataset taken from the “datasets” R package. Created here is a tree model based on it and saved in PMML. PMML files, input dataset and a complete C# sample illustrating this model is shipped with the installer.

    Samples location:

    %LOCALAPPDATA%\Syncfusion\EssentialStudio\%version%\Common\Analytics\Tree Model\Iris

    The following code example illustrates the procedure to call TreeModel evaluator directly without the help of PMMLEvaluatorFactory.

  • C#
    • //Sample values are passed as anonymous type            
var anonymousType = new            
{                
SepalLength = 6.2,                
SepalWidth = 2.9,                
PetalLength = 4.3,                
PetalWidth = 1.3            
};            
string pmmlFilePath = "../../Iris.pmml";  
          
//Create instance for PMMLDocument            
PMMLDocument pmmlDocument = new PMMLDocument(pmmlFilePath); 
           
//Create instance for TreeModel Evaluator            
TreeModelEvaluator treeEvaluator = new TreeModelEvaluator(pmmlDocument);
           
//Gets the predicted result            
PredictedResult predictedResult = treeEvaluator.GetResult(anonymousType, null);            
treeEvaluator.Dispose();            

//Displays the predicted result            
Console.WriteLine(predictedResult.PredictedValue);           


//Output

Output: Versicolor
The predicted output (Versicolor) obtained from PMML Execution engine gives us a clear picture that based on given sepal and petal width and length, the iris species is more likely to be Versicolor.

    Support Vector Machine Model Evaluator

    Support Vector Machine Model Evaluator represents the evaluator for support vector machine models and it inherits all the properties and methods of PMMLEvaluator.

    Support Vector Machines

    Support Vector Machines (SVM) model is a set of related supervised learning model with associated learning algorithms that analyze data and recognize patterns, used for both classification and regression analysis. It has gained popularity due to its potential for high accuracy.

    SupportVectorMachine Model Public Properties/Methods

    Property/Method Description
    SupportVectorMachineModelEvaluator (PMMLDocument) Creates Instance for SupportVectorMachine model evaluator for the given PMML document object.
    GetResult(object,IModelOptions) Evaluates the given input against the scoring procedure of the SupportVectorMachine model and returns the PredictedResult.Currently there is no Model options available for Support vector machine Model.

    Consider the Iris dataset taken from the “datasets” R package. Created here is a Support Vector Machine model based on it and saved in PMML. PMML files, input dataset and a complete C# sample illustrating this model is shipped with the installer.

    Samples location:

    %LOCALAPPDATA%\Syncfusion\EssentialStudio\%version%\Common\Analytics\Support Vector Machine\Iris Sigmoid

    The following code example illustrates the procedure to call Support Vector Machine model evaluator directly without the help of PMMLEvaluatorFactory.

  • C#
    • //Sample values are passed as anonymous type            
var anonymousType = new            
{               
SepalLength = 5.1,               
SepalWidth = 3.8,                
PetalLength = 1.6,                
PetalWidth = 0.2            
};            
string pmmlFilePath = "../../Iris Sigmoid.pmml";           

//Create instance for PMML Document            
PMMLDocument pmmlDocument = new PMMLDocument(pmmlFilePath);           

//Create instance for SupportVectorMachine model            
SupportVectorMachineModelEvaluator supportVector = new SupportVectorMachineModelEvaluator(pmmlDocument);            
 
//Gets the predicted result           
PredictedResult predictedResult = supportVector.GetResult(anonymousType, null);           
supportVector.Dispose();            
 
//Displays the predicted result            
Console.WriteLine(predictedResult.PredictedValue);


//Output

Output: Setosa
The predicted output (Setosa) obtained from PMML Execution engine gives us a clear picture that based on given sepal and petal - width and length, the iris species is more likely to be Setosa.

    Mining Model or Random Forest Evaluator

    Mining Model Evaluator represents the evaluator for mining models and it inherits all the properties and methods of PMMLEvaluator.

    Mining Model

    Mining model operates by constructing multiple decision trees. It draws random samples from the original data and for each random sample it grows a tree. It is used in both classification and regression cases.

    Mining Model Public Properties or Methods

    Property/Method Description
    MiningModelEvaluator (PMMLDocument) Creates Instance for MiningModel evaluator for the given input PMML document object.
    GetResult(object,IModelOptions) Evaluates the given input against the scoring procedure of the MiningModel and returns the PredictedResult.Currently there is no Model options available for Mining Model.

    NOTE

    Gradient Boosting Model also initializes the Mining Model Evaluator as the PMML structure is similar.

    Consider the Iris dataset taken from the “datasets” R package. Created here is a Random Forest model based on it and saved in PMML. PMML files, input dataset and a complete C# sample illustrating this model is shipped with the installer.

    Samples location:

    %LOCALAPPDATA%\Syncfusion\EssentialStudio\%version%\Common\Analytics\Random Forest\Iris

    The following code example illustrates the procedure to call Mining Model Evaluator directly without the help of PMMLEvaluatorFactory.

  • C#
    • //Sample values are passed as anonymous type           
var anonymousType = new            
{                
SepalLength = 6.8,                
SepalWidth = 3,               
PetalLength = 5.5,               
PetalWidth = 2.1            
};            
string pmmlFilePath = "../../Iris.pmml";            
 
//Create instance for PMML Document            
PMMLDocument pmmlDocument = new PMMLDocument(pmmlFilePath);            
 
//Create instance for Mining model            
MiningModelEvaluator miningModel = new MiningModelEvaluator(pmmlDocument);            
 
//Gets the predicted result            
PredictedResult predictedResult = miningModel.GetResult(anonymousType, null);
miningModel.Dispose();            
 
//Displays the predicted result            
Console.WriteLine(predictedResult.PredictedValue);


//Output

Output: Virginica
The predicted output (Virginica) obtained from PMML Execution engine gives us a clear picture that based on given sepal and petal width and length, the iris species is more likely to be Virginica.

    Neural Network Model Evaluator

    Neural Network Model Evaluator represents the evaluator for neural network models and it inherits all the properties and methods of PMMLEvaluator.

    Neural Network Model

    A Neural network also called an ANN or an Artificial Neural Network is an artificial system made of artificial neuron cells. It is modeled after the way the human brain works through imitating how the brain’s neurons are fired or activated. The idea of neural networks is that, they are able to learn by themselves, an ability that makes them remarkably distinctive in comparison to normal computers, that cannot do anything for that they are not programmed. It is used in both classification and regression cases.

    Neural Network Model Public Properties or Methods

    Property/Method Description
    NeuralNetworkModelEvaluator (PMMLDocument) Creates Instance for NeuralNetworkModel evaluator for the given input PMML document object.
    GetResult(object,IModelOptions) Evaluates the given input against the scoring procedure of the NeuralNetworkModel and returns the PredictedResult. Currently there is no Model options available for Neural Network Model.

    Consider the Iris dataset taken from the “datasets” R package. Created here is a Neural Network model based on it and saved in PMML. PMML files, input dataset and a complete C# sample illustrating this model is shipped with the installer.

    Samples location:

    %LOCALAPPDATA%\Syncfusion\EssentialStudio\%version%\Common\Analytics\Neural Networks\Iris

    The following code example illustrates the procedure to call Neural Network Model Evaluator directly without the help of PMMLEvaluatorFactory.

  • C#
    • //Sample values are passed as anonymous type            
var anonymousType = new            
{                
SepalLength = 5.2,                
SepalWidth = 3.3,                
PetalLength = 4.5,               
PetalWidth = 1.2            
};            
string pmmlFilePath = "../../Iris.pmml";            

//Create instance for PMML Document            
PMMLDocument pmmlDocument = new PMMLDocument(pmmlFilePath);            

//Create instance for Mining model            
NeuralNetworkModelEvaluator neuralNetworkModel = new NeuralNetworkModelEvaluator(pmmlDocument);            

//Gets the predicted result            
PredictedResult predictedResult = neuralNetworkModel.GetResult(anonymousType, null);
neuralNetworkModel.Dispose();            

//Displays the predicted result            
Console.WriteLine(predictedResult.PredictedValue);


//Output

Output: Versicolor
The predicted output (Versicolor) obtained from PMML Execution engine provides a clear picture that is based on given sepal and petal width and length, the iris species is more likely to be Versicolor.

    Clustering Model Evaluator

    Clustering Model Evaluator represents the evaluator for clustering models and it inherits all the properties and methods of PMMLEvaluator.

    Clustering Model

    Clustering is the task of grouping a set of objects in such a way that objects in the same group called a cluster are more similar in some sense or another to each other than to those in other groups (clusters). For each cluster a center vector can be provided. In center-based models a cluster is defined by a vector of center coordinates. Some distance measure is used to determine the nearest center that is the nearest cluster for a given input record.

    Clustering Model Public Properties or Methods

    Property/Method Description
    ClusteringModelEvaluator (PMMLDocument) Creates Instance for ClusteringModel evaluator for the given input PMML document object.
    GetResult(object,IModelOptions) Evaluates the given input against the scoring procedure of the ClusteringModel and returns the PredictedResult. Currently there is no Model options available for Clustering Model.

    Consider the Iris dataset taken from the “datasets” R package. Created here is a Clustering model based on it and saved in PMML. PMML files, input dataset and a complete C# sample illustrating this model is shipped with the installer.

    Samples location:

    %LOCALAPPDATA%\Syncfusion\EssentialStudio\%version%\Common\Analytics\Clustering Model\Iris

    The following code example illustrates the procedure to call Clustering Model Evaluator directly without the help of PMMLEvaluatorFactory.

  • C#
    • //Sample values are passed as anonymous type            
var anonymousType = new            
{                
SepalLength = 5.4,                
SepalWidth = 3.7,                
PetalLength = 3.5,                
PetalWidth = 1.8            
};            
string pmmlFilePath = "../../Iris.pmml";           

//Create instance for PMML Document            
PMMLDocument pmmlDocument = new PMMLDocument(pmmlFilePath);            
 
//Create instance for Mining model            
ClusteringModelEvaluator clusteringModel = new ClusteringModelEvaluator (pmmlDocument);            
 
//Gets the predicted result            
PredictedResult predictedResult = clusteringModel.GetResult(anonymousType, null);
clusteringModel.Dispose();           

//Displays the predicted result            
Console.WriteLine(predictedResult.PredictedValue);


//Output

Output: 2
The predicted output (2) obtained from PMML Execution engine gives us a clear picture that based on given sepal and petal width and length, the iris species is more likely belongs to the Cluster 2.

    Association Rules Model Evaluator

    Association Rules Model Evaluator represents the evaluator for the Association rules model and it inherits all the properties and methods of PMMLEvaluator.

    Association Rules Model

    Association rule mining finds elements or attributes that frequently occur together—for example, products that are often bought together during a shopping session. Such information can be used to recommend products to shoppers, to place frequently bundled items together on store shelves. Machine learning methods for identifying associations among items in transactional data called market basket analysis in retail stores.

    Association Rules Model Public Properties or Methods

    Property/Method Description
    AssociationRulesModelEvaluator(PMMLDocument) Creates Instance for Association Rules model evaluator for the given PMML document object.
    GetResult(object,IModelOptions) Evaluates the given input items against the scoring procedure of the Association rules model and returns the recommendations, exclusive recommendations and rule associations.

    Consider the Groceries dataset taken from the “arules” R package. An Association rules model is created here, based on it and saved in PMML. PMML files, input dataset and a complete C# sample illustrating this model is shipped with the installer.

    Samples location:

    %LOCALAPPDATA%\Syncfusion\EssentialStudio\%version%\Common\Analytics\Association Rules\Groceries

    The following code example illustrates the procedure to call Association rules model evaluator directly without the help of PMMLEvaluatorFactory.

  • C#
    • //Sample input items is passed as list of string            
List<string> input = null;            
input.Add(“citrus fruit”);            
input.Add(“semi-finished bread”);            
input.Add(“margarine”);            
input.Add(“ready soups”);            
string pmmlFilePath = "../../Groceries.pmml";           

//Create instance for PMMLDocument            
PMMLDocument pmmlDocument = new PMMLDocument(pmmlFilePath);            
 
//Create instance for AssociationRulesModelEvaluator            
AssociationRulesModelEvaluator associationRules = new AssociationRulesModelEvaluator(pmmlDocument);            
 
//Gets the predicted result            
PredictedResult predictedResult = associationRules.GetResult(input, null);            
associationRules.Dispose();            
string[] recommendations = predictedResult.GetRecommendations();            
string[] exclusiveRecommendations = predictedResult.GetExclusiveRecommendations();            
string[] ruleAssociations = predictedResult.GetRuleAssociations();            
 
//Displays the recommended results            
Console.WriteLine(“Recommendations: [” + string.Join(“,”,recommendations) + “]”);            
Console.WriteLine(“ExclusiveRecommendations: [” + string.Join(“,”,exclusiveRecommendations) + “]”);           
Console.WriteLine(“RuleAssociations: [” + string.Join(“,”,ruleAssociations) + “]”);


//Output

Output: Recommendations: [whole milk,rolls/buns,other vegetables,yogurt]
ExclusiveRecommendations: [whole milk,rolls/buns,other vegetables,yogurt]
RuleAssociations: []
The Recommendations, ExclusiveRecommendations and RuleAssociations above obtained from PMML Execution engine gives us a clear picture that based on given input items association rules suggests the recommendations that are often bought together during a shopping session.

    K-Nearest Neighbors Model Evaluator

    K-Nearest Neighbors Model Evaluator represents the evaluator for the K-Nearest Neighbors model and it inherits all the properties and methods of PMMLEvaluator.

    K-Nearest Neighbors Model

    K-Nearest Neighbors Model is a type of instance-based learning or lazy learning used for both classification and regression.

    Property/Method Description
    NearestNeighborModelEvaluator(PMMLDocument) Creates Instance for K-Nearest Neighbors model evaluator for the given PMML document object.
    GetResult(object,IModelOptions) Evaluates the given input items against the scoring procedure of the NearestNeighborModel and returns the PredictedResult. Currently there is no Model options available for K-Nearest Neighbors Model.

    Consider the Tips dataset taken from the “reshape2” R package that has information about each tip received by a waiter over a period of few months.

    The following code example illustrates the procedure to call K-Nearest Neighbor model evaluator directly without the help of PMMLEvaluatorFactory.

  • C#
    • //Sample values are passed as anonymous type          
var anonymousType = new
{		
total_bill = 16.99,
sex = "Female",
smoker = "No",
day = "Sun",
timing = "Dinner",
size = 2
};
string pmmlFilePath = "../../Tips.pmml";

//Create instance for PMML Document            
PMMLDocument pmmlDocument = new PMMLDocument(pmmlFilePath);

//Create instance for K-Nearest Neighbor model            
NearestNeighborModelEvaluator knnModel = new NearestNeighborModelEvaluator(pmmlDocument);

//Gets the predicted result            
PredictedResult predictedResult = knnModel.GetResult(anonymousType, null);
knnModel.Dispose();

//Displays the predicted result            
Console.WriteLine(predictedResult.PredictedValue);


//Output

Output: 2.67
The predicted output (2.67) obtained from PMML Execution engine gives us a clear picture that based on given information (independent variables), the waiter may get a tip of $2.67. Here we considered the independent variables(total_bill,sex,smoker,day,time,size) and by using these independent variables we can find the result of dependent variable (tip) which is mentioned above.

    PMMLDocument

    The PMMLDocument class is used to represent the object model for PMML document. We can load the input PMML file in both constructor as well as the open methods. Currently all properties provided with the object model are read only and allows us to check the values represented in the input PMML (XML) file.

    Properties and Methods

    The PMMLDocument contains properties with read only option.

    PMMLDocument Properties

    Properties/Methods Description
    Header Gets the Header values
    DataDictionary Gets the DataDictionary values
    SupportVectorMachineModel Gets the SupportVectorMachineModel elements values
    TreeModel Gets the TreeModel values
    RegressionModel Gets the RegressionModel values
    GeneralRegressionModel Gets the GeneralRegressionModel values
    NaiveBayesModel Gets the NaiveBayesModel values
    MiningModel Gets the MiningModel values
    NeuralNetworkModel Gets the NeuralNetworkModel values
    ClusteringModel Gets the ClusteringModel values
    AssociationRulesModel Gets the AssociationRulesModel values
    NearestNeighborModel Gets the NearestNeighborModel values
    PMMLDocument Initializes the class PMMLDocument
    PMMLDocument(string) Initializes PMMLDocument with string parameter
    PMMLDocument(Stream) Initializes PMMLDocument with stream parameter
    PMMLDocument(TextReader) Initializes PMMLDocument with text reader parameter
    OpenPMMLDocument(string) Opens the input PMML string file path
    OpenPMMLDocument(string, PMMLValidationType) Opens the input PMML string file path and performs schema validation based on the specified validation type
    OpenPMMLDocument(Stream) Opens the input PMML stream file path
    OpenPMMLDocument(Stream, PMMLValidationType) Opens the input PMML stream file and performs schema validation based on the specified validation type
    OpenPMMLDocument(TextReader) Opens the input PMML as text reader object
    OpenPMMLDocument(TextReader, PMMLValidationType) Opens the input PMML as text reader object and performs schema validation based on the specified validation type
    Dispose() Releases the memory occupied by objects

    The following code example illustrates you on how the PMML file is loaded. Here the input file path is read in string format.

  • C#
    • //Gets the filepath of the PMML file as String

string PmmlFilePath = "../../Iris.pmml";

PMMLDocument pmmlDocument = new PMMLDocument();

pmmlDocument.OpenPMMLDocument(PmmlFilePath);


//Filepath directly passed to its instance 

string PmmlFilePath = "../../Iris.pmml";

PMMLDocument pmmlDocument = new PMMLDocument(PmmlFilePath);

pmmlDocument.Dispose();

    The following code example illustrates you the same method OpenPMMLDocument but the file path is in stream type.

  • C#
    • //Gets the Stream filepath of the PMML file 

string PmmlFilePath = "../../Iris.pmml";

PMMLDocument pmmlDocument = new PMMLDocument();

FileStream pmmlStream = File.Open(PmmlFilePath, FileMode.Open, FileAccess.Read);

pmmlDocument.OpenPMMLDocument(pmmlStream);


//Filestream directly passed to the Instance of PMMLDocument

string PmmlFilePath = "../../Iris.pmml";

FileStream pmmlStream = File.Open(PmmlFilePath, FileMode.Open, FileAccess.Read);

PMMLDocument pmmlDocument = new PMMLDocument(pmmlStream);

pmmlDocument.Dispose();

    The following code example illustrates you the same method OpenPMMLDocument but the file is in text reader type.

  • C#
    • //Gets the text of the PMML file 

string PmmlFilePath = "../../Iris.pmml";

PMMLDocument pmmlDocument = new PMMLDocument();

TextReader pmmlText = new StringReader(File.ReadAllText(PmmlFilePath));

pmmlDocument.OpenPMMLDocument(pmmlText);


//TextReader directly passed to the Instance of PMMLDocument

string PmmlFilePath = "../../Iris.pmml";

TextReader pmmlText = new StringReader(File.ReadAllText(PmmlFilePath));

PMMLDocument pmmlDocument = new PMMLDocument(pmmlText);

pmmlDocument.Dispose();

    PredictedResult

    The PredictedResult class is used to represent the predicted output for all PMMLEvaluator.

    Properties and Methods

    PredictedResult Properties

    Properties/Methods Description
    PredictedField Gets the Predicted field (Column) name
    PredictedDoubleValue Gets the predicted value as double data type. It returns predicted value only when predicting numeric field.
    PredictedStringValue Gets the predicted value as string data type. It returns predicted value only when predicting categorical field.
    PredictedValue Gets the predicted value. It returns the predicted value as object data type.Note: For association rules model, Object will return string array of exclusive recommended items.
    PredictedDataType Gets the predicted data type It’s an enumeration illustrating the return type of predicted value. Return type of predicted data type is one of the following: Double, String or Array
    GetPredictedProbability Gets the predicted probability for the input category
    GetPredictedProbabilities Get the predicted probabilities for all categories as Key value pair
    GetPredictedCategories Gets the predicted categories
    GetOutputFieldResults Gets the Output field evaluation results.
    GetRecommendations Gets the array of recommended items.(This method should be used only for AssociationRules Model)
    GetExclusiveRecommendations Gets the array of exclusive recommended items.(This method should be used only for AssociationRules Model)
    GetRuleAssociations Gets the array of Rules associated items.(This method should be used only for AssociationRules Model)

    TreeModelResult

    The TreeModelResult class is used to represent the predicted output of Tree Model with the following extended properties,

    TreeModelResult Properties

    Properties/Methods Description
    PredictedNodeID Gets the predicted node ID
    Path Gets the traverse path (node IDs) of predicted node

    The following code example illustrates you on how to retrieve the predicted node ID from the predicted result,

  • C#
    • //Sample values are passed as anonymous type            
var anonymousType = new            
{                
SepalLength = 6.2,                
SepalWidth = 2.9,                
PetalLength = 4.3,                
PetalWidth = 1.3            
};            
string pmmlFilePath = "../../Iris.pmml";  
          
//Create instance for PMMLDocument            
PMMLDocument pmmlDocument = new PMMLDocument(pmmlFilePath); 
           
//Create instance for TreeModel Evaluator            
TreeModelEvaluator treeEvaluator = new TreeModelEvaluator(pmmlDocument);
           
//Gets the predicted result            
PredictedResult predictedResult = treeEvaluator.GetResult(anonymousType, null);   

// Gets the tree model result
TreeModelResult treeModelResult = ((TreeModelResult)predictedResult);         
treeEvaluator.Dispose();            

//Displays the predicted node ID 
Console.WriteLine("Predicted Node: " +treeModelResult.PredictedNodeID);

//Displays the traverse path of predicted node
Console.WriteLine("Traverse Path: " +treeModelResult.Path);      


//Output

Output: Predicted Node: 6, Traverse Path: 1,3,6

    AssociationModelResult

    The AssociationModelResult class is used to represent the predicted output of Association Rules Model with the following extended properties/method,

    AssociationModelResult Properties

    Properties/Methods Description
    GetConfidences(RecommendationType) Get the recommended items and its confidence values

    The following code example illustrates you on how to retrieve the confidence values from the predicted result,

  • C#
    • //Sample input items is passed as list of string            
List<string> input = new List<string>();            
input.Add("citrus fruit");            
input.Add("semi-finished bread");            
input.Add("margarine");            
input.Add("ready soups");            
string pmmlFilePath = "../../Groceries.pmml";           

//Create instance for PMMLDocument            
PMMLDocument pmmlDocument = new PMMLDocument(pmmlFilePath);            
 
//Create instance for AssociationRulesModelEvaluator            
AssociationRulesModelEvaluator associationRules = new AssociationRulesModelEvaluator(pmmlDocument);            
 
//Gets the predicted result            
PredictedResult predictedResult = associationRules.GetResult(input, null); 

//Gets the Association rule model result
AssociationModelResult associationModelResult = (AssociationModelResult)predictedResult; 
associationRules.Dispose();            

Dictionary<string, decimal> recommendationConfidences = associationModelResult.GetConfidences(RecommendationType.Recommendation);
Dictionary<string, decimal> exclusiveRecommendationConfidences = associationModelResult.GetConfidences(RecommendationType.ExclusiveRecommendation);
Dictionary<string, decimal> ruleAssociationConfidences = associationModelResult.GetConfidences(RecommendationType.RuleAssociation);  

//Displays the recommended results            
Console.WriteLine(“Recommendations: [” + string.Join(", ", recommendationConfidences.Select(m => m.Key + ":" + m.Value).ToArray()) + “]”);            
Console.WriteLine(“ExclusiveRecommendations: [” + string.Join(", ", exclusiveRecommendationConfidences.Select(m => m.Key + ":" + m.Value).ToArray()) + “]”);           
Console.WriteLine(“RuleAssociations: [” + string.Join(", ", ruleAssociationConfidences.Select(m => m.Key + ":" + m.Value).ToArray()) + “]”);


//Output

Output: Recommendations: [whole milk:0.413194444444444, rolls/buns:0.251736111111111, other vegetables:0.348894348894349, yogurt:0.261670761670762]
ExclusiveRecommendations: [whole milk:0.413194444444444, rolls/buns:0.251736111111111, other vegetables:0.348894348894349, yogurt:0.261670761670762]
RuleAssociations: []

    Prediction Sensitivity in Binomial Classification

    Binary classification is used to classify cases into two categories/groups. You are provided the option to increase or decrease the sensitivity of the binary prediction in the following models.

    1. General Regression model
    2. Naive Bayes model
    3. Neural Networks model

    Using BinomialThreshold property in Model options of General Regression, that is GeneralRegressionOptions, and Naive Bayes, that is NaiveBayesOptions, and Neural Networks, that is NeuralNetworkOptions you can change the threshold value between 0 to 1 as a threshold to increase or decrease the sensitivity of the binary prediction.

    Based on the value, probability (0 - 1) is split into two boundary regions.

    Binary Prediction

    Category 1 0 to BinomialThreshold
    Category 2 BinomialThreshold to 1

    The following code example illustrates the procedure to call General Regression Model Evaluator by passing the binomial threshold value in GeneralRegressionOptions.

  • C#
    • var iris = new            
{                
Sepal_Length = 6.5, 
Sepal_Width = 2.8, Petal_Length = 4.6,
Petal_Width = 1.5            
};            
var PMMLEvaluator = new PMMLEvaluatorFactory().GetPMMLEvaluatorInstance("../../Model/Iris Logit.pmml");            
GeneralRegressionOptions generalRegressionOptions = new GeneralRegressionOptions();            
generalRegressionOptions.BinomialThreshold = 0.8;            

//Get predicted result with Binomial Threshold = 0.8            
PredictedResult predictedResult = PMMLEvaluator.GetResult(iris, generalRegressionOptions);           
string[] fieldNames = predictedResult.GetPredictedCategories();            
for (int i = 0; i < fieldNames.Length; i++)            
{               
Console.WriteLine(string.Format("The probability of {0}({1}) is {2}",  fieldNames[i], fieldNames[i] == "0" ? "not versicolor" : "versicolor",
predictedResult.GetPredictedProbability(fieldNames[i])));     
}            
Console.WriteLine(string.Format("The PredictedResult with Binomial threshold value (0.8) is {0} ({1})",               
predictedResult.PredictedValue, predictedResult.PredictedValue.ToString() == "0" ? "not versicolor" : "versicolor"));           
generalRegressionOptions.BinomialThreshold = 0.2;            
 
//Get predicted result with Binomial Threshold = 0.2            
predictedResult = PMMLEvaluator.GetResult(iris, generalRegressionOptions);            
Console.WriteLine(string.Format("The PredictedResult with Binomial threshold value (0.2) is {0} ({1})",               
predictedResult.PredictedValue, predictedResult.PredictedValue.ToString() == "0" ? "not versicolor" : "versicolor"));


//Output

Output: The probability of 0 (not versicolor) is 0.424140583239206
The probability of 1 (versicolor) is 0.575859416760794
The PredictedResult with Binomial threshold value (0.8) is 0 (not versicolor) and the PredictedResult with Binomial threshold value (0.2) is 1 (versicolor).
When threshold value = 0.8, the probability value of 1 (versicolor) is 0.5758 which is less than our binomial threshold value (0.8) so the predicted result with binomial threshold value 0.8 is 0 (i.e.) not versicolor.
When threshold value = 0.2, the probability value of 1 (versicolor) is 0.5758 which is greater than our binomial threshold value (0.2) so the predicted result with binomial threshold value 0.2 is 1 (i.e.) versicolor.