/*
* Create mining model for the selected mining strucutre
*/
private void CreateModels(MiningStructure objStructure)
{
MiningModel ClusterModel;
MiningModel TreeModel;
MiningModelColumn mmc;
// Create the Cluster model and set the algorithm
// and parameters
DropExistingMiningModels(objStructure, sModelName);
ClusterModel = objStructure.CreateMiningModel(true, sModelName);
ClusterModel.Algorithm = MiningModelAlgorithms.MicrosoftClustering;// "Microsoft_Clustering";
ClusterModel.AlgorithmParameters.Add("CLUSTER_COUNT", 0);
ClusterModel.Update();
// The CreateMiningModel method adds
// all the structure columns to the collection
// Copy the Cluster model and change the necessary properties
TreeModel = ClusterModel.Clone();
DropExistingMiningModels(objStructure, sModelName + "Generation Trees");
TreeModel.Name = sModelName + "Generation Trees";
TreeModel.ID = sModelName + "Generation Trees";
TreeModel.Algorithm = MiningModelAlgorithms.MicrosoftDecisionTrees;// "Microsoft_Decision_Trees";
TreeModel.AlgorithmParameters.Clear();
TreeModel.Columns["Gender"].Usage = "Predict";
//TreeModel.Columns["PayChannels"].Usage = "Predict";
// Add an aliased copy of the PayChannels table to the trees model
mmc = TreeModel.Columns.Add("MaritalStatus");
mmc.SourceColumnID = "MaritalStatus";
mmc = mmc.Columns.Add("MaritalStatus");
mmc.SourceColumnID = "MaritalStatus";
mmc.Usage = "Key";
// Now set a filter on the PayChannels_Hbo_Encore table and use it
// as input to predict other channels
//TreeModel.Columns["PayChannels_Hbo_Encore"].Filter = "Channel=’HBO’ OR Channel=’Encore’";
// Set a complementary filter on the payChannels predictable
// nested table
//TreeModel.Columns["PayChannels"].Filter = "Channel<>’HBO’ AND Channel<>’Encore’";
objStructure.MiningModels.Add(TreeModel);
// Submit the models to the server
// ToDo: fix this
//TreeModel.Update();
}
/*
* Create mining model with custom fields and algorithm
*/
private void CreateCustomModel(MiningStructure objStructure, string sAlgorithm, string sModelName, string sKeyColumn, List <string> lPredictColumns, List <bool> lbPredictColumns, int parOne, int parTwo)
{
// drop existing model
if (objStructure.MiningModels.ContainsName(sModelName))
{
objStructure.MiningModels[sModelName].Drop();
}
// Detailed description of the model algorithms is here:
// http://msdn.microsoft.com/en-us/library/ms175595.aspx
// More customisation for these algorithms can be found here:
// http://msdn.microsoft.com/en-us/library/cc280427.aspx
// Also a model example can be found here:
// http://msdn.microsoft.com/en-us/library/ms345087(v=SQL.100).aspx
Microsoft.AnalysisServices.MiningModel myMiningModel = objStructure.CreateMiningModel(true, sModelName);
myMiningModel.Algorithm = sAlgorithm;
switch (sAlgorithm)
{
case MiningModelAlgorithms.MicrosoftClustering:
myMiningModel.AlgorithmParameters.Add("CLUSTERING_METHOD", parOne);
myMiningModel.AlgorithmParameters.Add("CLUSTER_COUNT", parTwo);
break;
//case MiningModelAlgorithms.MicrosoftTimeSeries:
// myMiningModel.AlgorithmParameters.Add("PERIODICITY_HINT", "{12}"); // {12} represents the number of months for prediction
// break;
case MiningModelAlgorithms.MicrosoftNaiveBayes:
break;
case MiningModelAlgorithms.MicrosoftDecisionTrees:
myMiningModel.AlgorithmParameters.Add("SCORE_METHOD", parOne);
myMiningModel.AlgorithmParameters.Add("SPLIT_METHOD", parTwo);
break;
}
/***************** Predict columns *****************/
// add optional predict columns
if (lPredictColumns.Count != 0)
{
// predict columns
for (int i = 0; i < lPredictColumns.Count; i++)
{
Microsoft.AnalysisServices.MiningModelColumn modelColumn = myMiningModel.Columns.GetByName(lPredictColumns[i]);
modelColumn.SourceColumnID = lPredictColumns[i];
if (lbPredictColumns[i] == true)
{
modelColumn.Usage = MiningModelColumnUsages.PredictOnly;
}
else
{
modelColumn.Usage = MiningModelColumnUsages.Predict;
}
}
}
myMiningModel.Update();
}
/*
* Create mining model with custom fields and algorithm
*/
private void CreateCustomModel(MiningStructure objStructure, string sAlgorithm, string sModelName, string sKeyColumn, List<string> lPredictColumns, List<bool> lbPredictColumns, int parOne, int parTwo)
{
// drop existing model
if (objStructure.MiningModels.ContainsName(sModelName))
objStructure.MiningModels[sModelName].Drop();
// Detailed description of the model algorithms is here:
// http://msdn.microsoft.com/en-us/library/ms175595.aspx
// More customisation for these algorithms can be found here:
// http://msdn.microsoft.com/en-us/library/cc280427.aspx
// Also a model example can be found here:
// http://msdn.microsoft.com/en-us/library/ms345087(v=SQL.100).aspx
Microsoft.AnalysisServices.MiningModel myMiningModel = objStructure.CreateMiningModel(true, sModelName);
myMiningModel.Algorithm = sAlgorithm;
switch (sAlgorithm)
{
case MiningModelAlgorithms.MicrosoftClustering:
myMiningModel.AlgorithmParameters.Add("CLUSTERING_METHOD", parOne);
myMiningModel.AlgorithmParameters.Add("CLUSTER_COUNT", parTwo);
break;
//case MiningModelAlgorithms.MicrosoftTimeSeries:
// myMiningModel.AlgorithmParameters.Add("PERIODICITY_HINT", "{12}"); // {12} represents the number of months for prediction
// break;
case MiningModelAlgorithms.MicrosoftNaiveBayes:
break;
case MiningModelAlgorithms.MicrosoftDecisionTrees:
myMiningModel.AlgorithmParameters.Add("SCORE_METHOD", parOne);
myMiningModel.AlgorithmParameters.Add("SPLIT_METHOD", parTwo);
break;
}
/***************** Predict columns *****************/
// add optional predict columns
if (lPredictColumns.Count != 0)
{
// predict columns
for (int i = 0; i < lPredictColumns.Count; i++)
{
Microsoft.AnalysisServices.MiningModelColumn modelColumn = myMiningModel.Columns.GetByName(lPredictColumns[i]);
modelColumn.SourceColumnID = lPredictColumns[i];
if (lbPredictColumns[i] == true)
modelColumn.Usage = MiningModelColumnUsages.PredictOnly;
else
modelColumn.Usage = MiningModelColumnUsages.Predict;
}
}
myMiningModel.Update();
}
/*
* Create mining model for the selected mining strucutre
*/
private void CreateModels(MiningStructure objStructure)
{
MiningModel ClusterModel;
MiningModel TreeModel;
MiningModelColumn mmc;
// Create the Cluster model and set the algorithm
// and parameters
DropExistingMiningModels(objStructure, sModelName);
ClusterModel = objStructure.CreateMiningModel(true, sModelName);
ClusterModel.Algorithm = MiningModelAlgorithms.MicrosoftClustering;// "Microsoft_Clustering";
ClusterModel.AlgorithmParameters.Add("CLUSTER_COUNT", 0);
ClusterModel.Update();
// The CreateMiningModel method adds
// all the structure columns to the collection
// Copy the Cluster model and change the necessary properties
TreeModel = ClusterModel.Clone();
DropExistingMiningModels(objStructure, sModelName + "Generation Trees");
TreeModel.Name = sModelName + "Generation Trees";
TreeModel.ID = sModelName + "Generation Trees";
TreeModel.Algorithm = MiningModelAlgorithms.MicrosoftDecisionTrees;// "Microsoft_Decision_Trees";
TreeModel.AlgorithmParameters.Clear();
TreeModel.Columns["Gender"].Usage = "Predict";
//TreeModel.Columns["PayChannels"].Usage = "Predict";
// Add an aliased copy of the PayChannels table to the trees model
mmc = TreeModel.Columns.Add("MaritalStatus");
mmc.SourceColumnID = "MaritalStatus";
mmc = mmc.Columns.Add("MaritalStatus");
mmc.SourceColumnID = "MaritalStatus";
mmc.Usage = "Key";
// Now set a filter on the PayChannels_Hbo_Encore table and use it
// as input to predict other channels
//TreeModel.Columns["PayChannels_Hbo_Encore"].Filter = "Channel=’HBO’ OR Channel=’Encore’";
// Set a complementary filter on the payChannels predictable
// nested table
//TreeModel.Columns["PayChannels"].Filter = "Channel<>’HBO’ AND Channel<>’Encore’";
objStructure.MiningModels.Add(TreeModel);
// Submit the models to the server
// ToDo: fix this
//TreeModel.Update();
}
/// <summary>
/// Generate a mining structure and full process it
/// </summary>
/// <param name="objDatabase">Analysis Service Database instance</param>
/// <param name="objDataSourceView">Analysis Service DataSourceView instance</param>
/// <param name="strCaseTableName">Mining table name</param>
/// <returns>Mining structure</returns>
public static object GenerateMiningStructure(Database objDatabase, DataSourceView objDataSourceView, string strCaseTableName, DecisionTreeAlgorithmParameters dtParams)
{
try
{
MiningStructure objMiningStructure = new MiningStructure();
objMiningStructure = objDatabase.MiningStructures.Add(objDatabase.MiningStructures.GetNewName(StringEncode(strCaseTableName)));
objMiningStructure.HoldoutMaxPercent = dtParams.HoldoutMaxPercent; // Percent for testing
objMiningStructure.Source = new DataSourceViewBinding(objDataSourceView.ID);
objMiningStructure.CaseTableName = strCaseTableName;
foreach (string name in getAllColumnName(objDataSourceView, strCaseTableName))
{
string colName = StringEncode(name);
ScalarMiningStructureColumn column = new ScalarMiningStructureColumn(colName, colName);
switch (colName)
{
case "ID":
// ProfileBasicId column
column.Type = MiningStructureColumnTypes.Long;
column.Content = MiningStructureColumnContents.Key;
column.IsKey = true;
// Add data binding to the column
column.KeyColumns.Add(strCaseTableName, name);
// Add the column to the mining structure
objMiningStructure.Columns.Add(column);
break;
case "ProfileBasicId":
case "JobPostingId":
case "UserId":
case "JobTitle":
case "JobName":
case "CompanyId":
case "CompanyName":
//column.Type = MiningStructureColumnTypes.Text;
//column.Content = MiningStructureColumnContents.Discrete;
break;
case "IsApproved":
default:
column.Type = MiningStructureColumnTypes.Boolean;
column.Content = MiningStructureColumnContents.Discrete;
// Add data binding to the column
column.KeyColumns.Add(strCaseTableName, name);
// Add the column to the mining structure
objMiningStructure.Columns.Add(column);
break;
}
}
MiningModel objMiningModel = objMiningStructure.CreateMiningModel(true, StringEncode(strCaseTableName));
//MiningModel objMiningModel = objMiningStructure.MiningModels.Add(objMiningStructure.MiningModels.GetNewName(strMiningStructureName));
objMiningModel.Algorithm = MiningModelAlgorithms.MicrosoftDecisionTrees;
objMiningModel.AllowDrillThrough = true;
objMiningModel.AlgorithmParameters.Add("SCORE_METHOD", dtParams.SCORE_METHOD);
objMiningModel.AlgorithmParameters.Add("COMPLEXITY_PENALTY", dtParams.COMPLEXITY_PENALTY);
objMiningModel.AlgorithmParameters.Add("SPLIT_METHOD", dtParams.SPLIT_METHOD);
objMiningModel.AlgorithmParameters.Add("MAXIMUM_INPUT_ATTRIBUTES", dtParams.MAXIMUM_INPUT_ATTRIBUTES);
objMiningModel.AlgorithmParameters.Add("MAXIMUM_OUTPUT_ATTRIBUTES", dtParams.MAXIMUM_OUTPUT_ATTRIBUTES);
objMiningModel.AlgorithmParameters.Add("MINIMUM_SUPPORT", dtParams.MINIMUM_SUPPORT);
int i = 0;
foreach(MiningModelColumn col in objMiningModel.Columns)
{
switch (col.Name)
{
case "IsApproved":
objMiningModel.Columns[i].Usage = "Predict";
break;
case "ID":
objMiningModel.Columns[i].Usage = "Key";
break;
default:
objMiningModel.Columns[i].Usage = "Input";
break;
}
++i;
}
//objMiningModel.Update(UpdateOptions.ExpandFull);
objMiningStructure.Update(UpdateOptions.ExpandFull);
Console.WriteLine("Processing mining model " + objMiningStructure.Name + "...");
objMiningModel.Process(ProcessType.ProcessFull);
Console.WriteLine("Process " + objMiningStructure.Name + " finished!");
return objMiningStructure;
}
catch (Exception ex)
{
throw new Exception("Error in Creating a Mining structure - GenerateMiningStructure. Error Message -> " + ex.Message);
}
}
