private Patient.Classification PerfectMatch()
{
//getting index of selected input patient
string input = txtNum.Text;
int inputId = Convert.ToInt32(input);
int inputIndex = inputList.FindIndex(x => x.ID == inputId);
Patient.Classification classification = Patient.Classification.Unknown;
for (int data = 0; data < dataList.Count; data++)
{
if ((dataList[data].GetAllAttributes().Where(entry => inputList[inputIndex].GetAllAttributes()[entry.Key] != entry.Value).ToDictionary(entry => entry.Key, entry => entry.Value)).Count == 0)
{
perfectMatchLabel.Visibility = Visibility.Visible;
perfectMatch.Visibility = Visibility.Visible;
if (typeof(Patient.Classification).IsEnumDefined(Convert.ToString(dataList[data].Diagnosis)))
{
inputList[inputIndex].Diagnosis = (Patient.Classification)Enum.Parse(typeof(Patient.Classification), Convert.ToString(dataList[data].Diagnosis));
}
classification = inputList[inputIndex].Diagnosis;
}
}
return(classification);
}
private Patient.Classification RecommendedClassiciation(Patient.Classification distanceClassification, Patient.Classification perfectMatch, List <Patient.Classification> weka)
{
List <Patient.Classification> classifications = new List <Patient.Classification>
{
weka[0],
weka[1],
distanceClassification
};
if (perfectMatch != Patient.Classification.Unknown)
{
classifications.Add(perfectMatch);
}
var frequency = classifications.GroupBy(x => x).ToDictionary(x => x.Key, x => x.Count());
Patient.Classification recommendation;
//If no majority diagnosis, diagnose in order of severity.
if (frequency.Values.Distinct().Count() == 1 && frequency.ContainsValue(1))
{
if (classifications.Contains(Patient.Classification.Cancer))
{
recommendation = Patient.Classification.Cancer;
}
else if (classifications.Contains(Patient.Classification.Polyp))
{
recommendation = Patient.Classification.Polyp;
}
else if (classifications.Contains(Patient.Classification.Infection))
{
recommendation = Patient.Classification.Infection;
}
else if (classifications.Contains(Patient.Classification.Normal))
{
recommendation = Patient.Classification.Normal;
}
else
{
recommendation = Patient.Classification.Unknown;
}
}
else
{
//finds majority diagnosis
recommendation = (from item in classifications
group item by item into g
orderby g.Count() descending
select g.Key).First();
}
return(recommendation);
}
private void UpdateUI(Patient.Classification distanceClassification, Patient.Classification perfectMatchClassification, Patient.Classification j48Classification, Patient.Classification nngeClassification, Patient.Classification recommendation, Patient.Classification givenClassification)
{
DistanceResult.Content = distanceClassification;
j48Result.Content = j48Classification;
NNgeResult.Content = nngeClassification;
RecommendedDiagnosis.Content = recommendation;
perfectMatch.Content = perfectMatchClassification;
if (givenClassification != Patient.Classification.Unknown)
{
givenDiagnosis.Content = givenClassification;
}
}
public void SetClassification(string in_classification)
{
if (in_classification == "Cancer")
{
classification = Patient.Classification.Cancer;
}
else if (in_classification == "Polyp")
{
classification = Patient.Classification.Polyp;
}
else if (in_classification == "Normal")
{
classification = Patient.Classification.Normal;
}
else if (in_classification == "Infection")
{
classification = Patient.Classification.Infection;
}
}
private Patient.Classification RecommendedClassification(Patient.Classification distanceClassification, Patient.Classification perfectMatch, List <Patient.Classification> weka)
{
List <Patient.Classification> classifications = new List <Patient.Classification>
{
weka[0],
weka[1],
distanceClassification
};
if (perfectMatch != Patient.Classification.Unknown)
{
classifications.Add(perfectMatch);
}
Patient.Classification recommendation = MostFrequentClass(classifications);
inputList[int.Parse(txtNum.Text) - 1].Diagnosis = recommendation;
return(recommendation);
}
private void GenerateButton_Click(object sender, RoutedEventArgs e)
{
ClearUI();
classificationDistances.Clear();
Patient.Classification distanceClassification = WeightedEuclidClassify();
Patient.Classification perfectMatchClassification = PerfectMatch();
List <Patient.Classification> wekaClassifications = WekaClassify();
Patient.Classification recommendation = RecommendedClassiciation(distanceClassification, perfectMatchClassification, wekaClassifications);
Patient.Classification givenClassification = Patient.Classification.Unknown;
if (inputList[int.Parse(txtNum.Text) - 1].GivenDiagnosis != Patient.Classification.Unknown)
{
givenClassification = inputList[int.Parse(txtNum.Text) - 1].GivenDiagnosis;
givenDiagnosisLabel.Visibility = Visibility.Visible;
givenDiagnosis.Visibility = Visibility.Visible;
}
UpdateUI(distanceClassification, perfectMatchClassification, wekaClassifications[0], wekaClassifications[1], recommendation, givenClassification);
}
private Patient.Classification WeightedEuclidClassify()
{
List <Patient> cancerData = dataList.Where(C => C.Diagnosis == Patient.Classification.Cancer).ToList();
List <Patient> infectionData = dataList.Where(C => C.Diagnosis == Patient.Classification.Infection).ToList();
List <Patient> polypData = dataList.Where(C => C.Diagnosis == Patient.Classification.Polyp).ToList();
List <Patient> normalData = dataList.Where(C => C.Diagnosis == Patient.Classification.Normal).ToList();
List <List <Patient> > completeData = new List <List <Patient> >
{
cancerData,
infectionData,
polypData,
normalData
};
double diff = 0;
double euclid = 0;
double weightedEuclid = 0;
//getting index of selected input patient
int inputId = Convert.ToInt32(txtNum.Text);
int inputIndex = inputList.FindIndex(x => x.ID == inputId);
foreach (var dataSet in completeData)
{
List <double> weightedDistances = new List <double>();
List <double> distances = new List <double>();
for (int data = 0; data < dataSet.Count; data++)
{
weightedEuclid = 0;
euclid = 0;
for (int attr = 1; attr < rankingList.Count; attr++)
{
foreach (var item in rankingList)
{
if (item.GetNumber() == attr)
{
double weight = item.GetWeight();
double inputVal = inputList[inputIndex].GetSingleAttribute(attr - 1);
double dataVal = dataSet[data].GetSingleAttribute(attr - 1);
diff = inputVal - dataVal;
euclid += Math.Pow(diff, 2);
weightedEuclid += weight * Math.Sqrt(Math.Pow(diff, 2));
break;
}
}
}
distances.Add(Math.Sqrt(euclid));
weightedDistances.Add(weightedEuclid);
}
//List<double> weightedDistances_2 = weightedDistances;
classificationDistances.Add(dataSet[0].Diagnosis, weightedDistances);
}
//getting smallest mean distance
double minMean = int.MaxValue;
Patient.Classification closestAverageClassification = new Patient.Classification();
foreach (var item in classificationDistances)
{
if (item.Value.Average() < minMean)
{
closestAverageClassification = item.Key;
minMean = item.Value.Average();
}
}
return(closestAverageClassification);
}
private List <Patient.Classification> WekaClassify()
{
//getting index of selected input patient
int inputId = Convert.ToInt32(txtNum.Text);
int inputIndex = inputList.FindIndex(x => x.ID == inputId);
Patient selectedPatient = inputList[inputIndex];
Patient.Classification j48Classification = Patient.Classification.Unknown;
Patient.Classification nngeClassification = Patient.Classification.Unknown;
//iterating through models
for (int modeli = 0; modeli < wekaModels.Count; modeli++)
{
//iterating through model rules
for (int rulei = 0; rulei < wekaModels[modeli].Rules.Count; rulei++)
{
bool match = false;
//iterating through ruleitems
for (int ruleItemi = 0; ruleItemi < wekaModels[modeli].Rules[rulei].RuleItems.Count; ruleItemi++)
{
RuleItem ruleItem = wekaModels[modeli].Rules[rulei].RuleItems[ruleItemi];
int attributeIndex = patientFeatures.IndexOf(ruleItem.Feature) - 1;
double patientValue = selectedPatient.GetSingleAttribute(attributeIndex);
List <double> ruleVals = ruleItem.GetValues();
List <RuleItem.Operators> ruleOps = ruleItem.GetOps();
for (int i = 0; i < ruleVals.Count; i++)
{
if (ruleOps[i] == RuleItem.Operators.Equal)
{
if (patientValue == ruleVals[i])
{
match = true;
}
else
{
match = false;
break;
}
}
else if (ruleOps[i] == RuleItem.Operators.Less)
{
if (patientValue < ruleVals[i])
{
match = true;
}
else
{
match = false;
break;
}
}
else if (ruleOps[i] == RuleItem.Operators.Greater)
{
if (patientValue > ruleVals[i])
{
match = true;
}
else
{
match = false;
break;
}
}
else if (ruleOps[i] == RuleItem.Operators.LessOrEqual)
{
if (patientValue <= ruleVals[i])
{
match = true;
}
else
{
match = false;
break;
}
}
else if (ruleOps[i] == RuleItem.Operators.GreatorOrEqual)
{
if (patientValue >= ruleVals[i])
{
match = true;
}
else
{
match = false;
break;
}
}
}
if (!match)
{
break;
}
}
if (match)
{
if (modeli == 0)
{
j48Classification = wekaModels[modeli].Rules[rulei].Classification;
}
else
{
nngeClassification = wekaModels[modeli].Rules[rulei].Classification;
}
break;
}
}
}
return(new List <Patient.Classification> {
j48Classification, nngeClassification
});
}
