#region License Information
/* HeuristicLab
* Copyright (C) 2002-2008 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
*
* This file is part of HeuristicLab.
*
* HeuristicLab is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* HeuristicLab is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with HeuristicLab. If not, see .
*/
#endregion
using System;
using System.Collections.Generic;
using System.Text;
using System.Xml;
using HeuristicLab.Core;
using HeuristicLab.Data;
using HeuristicLab.DataAnalysis;
using System.Linq;
namespace HeuristicLab.Modeling {
public class VariableImpactCalculator : OperatorBase {
public override string Description {
get { return @"Calculates the impact of all allowed input variables on the quality of the model using evaluator supplied as suboperator."; }
}
public VariableImpactCalculator()
: base() {
AddVariableInfo(new VariableInfo("Dataset", "Dataset", typeof(Dataset), VariableKind.In));
AddVariableInfo(new VariableInfo("TargetVariable", "TargetVariable", typeof(IntData), VariableKind.In));
AddVariableInfo(new VariableInfo("AllowedFeatures", "Indexes of allowed input variables", typeof(ItemList), VariableKind.In));
AddVariableInfo(new VariableInfo("TrainingSamplesStart", "TrainingSamplesStart", typeof(IntData), VariableKind.In));
AddVariableInfo(new VariableInfo("TrainingSamplesEnd", "TrainingSamplesEnd", typeof(IntData), VariableKind.In));
AddVariableInfo(new VariableInfo("VariableImpacts", "Variable impacts", typeof(ItemList), VariableKind.New));
}
public override IOperation Apply(IScope scope) {
ItemList allowedFeatures = GetVariableValue>("AllowedFeatures", scope, true);
int targetVariable = GetVariableValue("TargetVariable", scope, true).Data;
Dataset dataset = GetVariableValue("Dataset", scope, true);
Dataset dirtyDataset = (Dataset)dataset.Clone();
int start = GetVariableValue("TrainingSamplesStart", scope, true).Data;
int end = GetVariableValue("TrainingSamplesEnd", scope, true).Data;
if (SubOperators.Count < 1) throw new InvalidOperationException("VariableImpactCalculator needs a suboperator to evaluate the model");
IOperator evaluationOperator = this.SubOperators[0];
ItemList variableImpacts = new ItemList();
// calculateReferenceQuality
double referenceQuality = CalculateQuality(scope, dataset, evaluationOperator);
for (int i = 0; i < allowedFeatures.Count; i++) {
int currentVariable = allowedFeatures[i].Data;
var oldValues = ReplaceVariableValues(dirtyDataset, currentVariable , CalculateNewValues(dirtyDataset, currentVariable, start, end), start, end);
double newQuality = CalculateQuality(scope, dirtyDataset, evaluationOperator);
double ratio = referenceQuality / newQuality;
double impact = ratio < 1.0 ? 1.0 - ratio : 1.0 - 1.0 / ratio;
ItemList row = new ItemList();
row.Add(new StringData(dataset.GetVariableName(currentVariable)));
row.Add(new DoubleData(impact));
variableImpacts.Add(row);
ReplaceVariableValues(dirtyDataset, currentVariable, oldValues, start, end);
}
scope.AddVariable(new Variable(scope.TranslateName("VariableImpacts"), variableImpacts));
return null;
}
private double CalculateQuality(IScope scope, Dataset dataset, IOperator evaluationOperator) {
Scope s = new Scope();
s.AddVariable(new Variable("Dataset", dataset));
scope.AddSubScope(s);
evaluationOperator.Execute(s);
double quality = s.GetVariableValue("Quality", false).Data;
scope.RemoveSubScope(s);
return quality;
}
private IEnumerable ReplaceVariableValues(Dataset ds, int variableIndex, IEnumerable newValues, int start, int end) {
double[] oldValues = new double[end - start];
for (int i = 0; i < end - start; i++) oldValues[i] = ds.GetValue(i + start, variableIndex);
if (newValues.Count() != end - start) throw new ArgumentException("The length of the new values sequence doesn't match the required length (number of replaced values)");
int index = start;
foreach(double v in newValues) {
ds.SetValue(index++, variableIndex, v);
}
return oldValues;
}
private IEnumerable CalculateNewValues(Dataset ds, int variableIndex, int start, int end) {
double mean = ds.GetMean(variableIndex, start, end);
return Enumerable.Repeat(mean, end - start);
}
}
}