public static StatisticalResultCsv[] Run(string name, string method, int ke, int keep, IEnumerable <StatisticalResultCsv> selectedMembers)
{
string methodFolder = $"{Config.WorkingFolder}\\{name}";
Directory.CreateDirectory(methodFolder);
string candidatesFolder = $"{Config.WorkingFolder}\\{name}\\candidates\\";
Directory.CreateDirectory(candidatesFolder);
foreach (var member in selectedMembers)
{
File.Copy($"{Config.WorkingFolder}\\{member.Method}\\{member.Dataset}\\individuals\\{member.File}", $"{candidatesFolder}\\{member.File}", true);
}
Icas.Common.ProcessExtension.Run("RScript", $"job_ensemble.R {name} {method} {selectedMembers.First().Dataset} {selectedMembers.First().Degredome} {keep}");
FeatureType dt = FeatureType.Reactivity;
FeatureType.TryParse(selectedMembers.First().DataType, out dt);
ProcessExtension.Run("Python", $"{Config.JobOnewayPy} {name} {selectedMembers.First().Dataset}");
ProcessExtension.Run("RScript", $"{Config.JobPairwiseComparisonR} {name} {selectedMembers.First().Dataset}");
AlgorithmCsv algorithm = new AlgorithmCsv();
algorithm.Name = name;
algorithm.Feature = selectedMembers.First().DataType;
DatasetCsv dataset = MiSettings.Datasets.First(c => c.Name == selectedMembers.First().Dataset);
var results = Icas.Clustering.Cluster.RunIndividual(algorithm, new[] { dataset }, false);
Icas.Clustering.Cluster.Summarize(algorithm);
return(results);
}
public void Filter(AlgorithmCsv algorithm)
{
var all = MiSettings.Datasets;
var filtered = all.Where(c => c.Feature == algorithm.Feature);
if (!string.IsNullOrWhiteSpace(algorithm.Transformation))
{
filtered = filtered.Where(c => c.Transformation == algorithm.Transformation);
}
this.DataSource = filtered.ToArray();
}
public static StatisticalResultCsv[] AttachCompactness(AlgorithmCsv algorithm, DatasetCsv dataset)
{
List <StatisticalResultCsv> final_results = new List <StatisticalResultCsv>();
double best_sr = 0;
double min_compact = 0;
int min_compact_at = 0;
string current_folder = $"{Config.WorkingFolder}\\{algorithm.Name}\\{dataset.Name}\\";
var dataType = FeatureTypeExtension.FromString(dataset.Feature);
PairwiseComparisonCsv[] pairwiseResults = CsvContext.ReadFile <PairwiseComparisonCsv>(current_folder + Config.PairwiseComparisonCsv);
if (pairwiseResults.Length == 0)
{
return(OutputResults(final_results, algorithm.Name, dataset.Name));
}
best_sr = pairwiseResults[0].significant_rate;
min_compact = GetCompactness(current_folder + "\\individuals\\" + pairwiseResults[0].File, dataset);
PairwiseComparisonCsv[] best_pairwiseResults = pairwiseResults.Where(c => c.significant_rate == best_sr).ToArray();
for (int i = 0; i < best_pairwiseResults.Length; i++)
{
//double compact =
StatisticalResultCsv finalResult = new StatisticalResultCsv();
finalResult.DataType = dataType.ToString();
finalResult.Method = algorithm.Name;
DatasetCsvToFinalCsv(dataset, finalResult);
PairCsvToFinalCsv(best_pairwiseResults[i], finalResult);
finalResult.Compactness = GetCompactness(current_folder + "\\individuals\\" + best_pairwiseResults[i].File, dataset);
if (finalResult.Compactness < min_compact)
{
min_compact = finalResult.Compactness;
min_compact_at = i;
}
final_results.Add(finalResult);
}
final_results[min_compact_at].Best = true;
for (int i = best_pairwiseResults.Length; i < best_pairwiseResults.Length + 5 && i < pairwiseResults.Length; i++)
{
StatisticalResultCsv finalResult = new StatisticalResultCsv();
finalResult.DataType = dataType.ToString();
finalResult.Method = algorithm.Name;
DatasetCsvToFinalCsv(dataset, finalResult);
PairCsvToFinalCsv(pairwiseResults[i], finalResult);
finalResult.Compactness = GetCompactness(current_folder + "\\individuals\\" + pairwiseResults[i].File, dataset);
final_results.Add(finalResult);
}
return(OutputResults(final_results, algorithm.Name, dataset.Name));
}
public static void Summarize(AlgorithmCsv algorithm)
{
//string[] files = Directory.GetFiles($"{Config.WorkingFolder}\\results\\", "*_all.csv");
List <StatisticalResultCsv> results = new List <StatisticalResultCsv>();
string[] subFolders = Directory.GetDirectories($"{Config.WorkingFolder}\\{algorithm.Name}\\");
foreach (var subFolder in subFolders)
{
foreach (var file in Directory.GetFiles(subFolder))
{
var fi = new System.IO.FileInfo(file);
if (fi.Name.StartsWith(algorithm.Name) && fi.Name.EndsWith("_results.csv"))
{
results.AddRange(CsvContext.ReadFile <StatisticalResultCsv>(file));
}
}
}
CsvContext.WriteFile($"{Config.WorkingFolder}\\results\\{algorithm}_all.csv", results, CsvConfig.Default);
StatisticalResultCsv[] best_results = results.Where(c => c.Best).ToArray();
CsvContext.WriteFile($"{Config.WorkingFolder}\\results\\{algorithm}_summary.csv", best_results, CsvConfig.Default);
}
public static StatisticalResultCsv[] RunIndividual(AlgorithmCsv algorithm, DatasetCsv dataset, int k, bool runScript)
{
List <StatisticalResultCsv> finalResults = new List <StatisticalResultCsv>();
FeatureType dataType = FeatureTypeExtension.FromString(dataset.Feature);
if (runScript)
{
//ProcessExtension.Run("python", $"\"{Config.WorkingFolder}{algorithm.Script}\" {dataset.Name} 10 100");
ProcessExtension.RunScript($"{Config.WorkingFolder}{algorithm.Script}", $"{dataset.Name} 10 100 {k}");
//onewan ANOVA
ProcessExtension.Run("python", $"\"{Config.WorkingFolder}{Config.JobOnewayPy}\" {algorithm.Name} {dataset.Name}");
//pairwise comparison
ProcessExtension.Run("RScript", $"\"{Config.WorkingFolder}{Config.JobPairwiseComparisonR}\" {algorithm.Name} {dataset.Name}");
}
//compactness
string folder = $"{Config.WorkingFolder}{algorithm.Name}\\{dataset.Name}\\";
if (!Directory.Exists(folder))
{
throw new FileNotFoundException();
}
switch (dataType)
{
case FeatureType.Reactivity:
finalResults.AddRange(CSCompactness.AttachCompactness(algorithm, dataset));
break;
case FeatureType.RnaDistance:
finalResults.AddRange(CSCompactness.AttachCompactness(algorithm, dataset));
break;
}
return(finalResults.ToArray());
}
