void process01EntropyAlgo()
{
this.dt_ready = TheTool.dataTable_partitize(dt_raw_exclude, partition_range); //01 partitize
//------------------------------------------------------------------------------
this.EnAlgo_first_list_F = TheTool.dataTable_getColList(dt_raw_exclude); //unselected feature (pointed by id)
this.list_entropy = TheEntropy.calEntropy(dt_ready);
this.list_MI = TheEntropy.calMIList(dt_ready, list_entropy); //for reuse MI
//Feature Selection Start : get first --------------------------------------
String fi1 = TheEntropy.temp_colWithHigestEntropy;
this.EnAlgo_first_list_F.Remove(fi1);
this.EnAlgo_first_S = TheData.getFeature(fi1, dt_threshold_pose1, dt_threshold_pose2);//Initial Threshold is assigned here (Not optimized)
}
void process01DCFS()
{
//prepare
this.DCFS_listDelta = getListDelta();
this.DCFS_listDeltaCorrelation = TheTool.list_Copy(DCFS_listDelta);
this.DCFS_first_list_F = TheTool.dataTable_getColList(dt_raw_exclude);
this.DCFS_listCorrelation = getlistCorrelation();
//pick first feature
string fi1 = DCFS_listDelta_getMaxValue();
this.DCFS_first_list_F.Remove(fi1);
this.DCFS_first_S = TheData.getFeature(fi1, dt_threshold_pose1, dt_threshold_pose2);
DCFS_listDelta_update(fi1);
}
//path_prefix = "" if no export needed
//looptime is negative in case auto
public List <Feature> process02_selectCandidate(string path_prefix, int methodSelection, int loopTime)
{
// F: threshold not needed, so just string
// S: threshold is needed, so use Feature
this.score_regression_summary = new List <String>();//keep summary
String selected;
List <string> local_list_F = new List <string>();
List <Feature> local_list_S0 = new List <Feature>();
List <string> local_list_S = new List <string>();
List <m_If> local_rules_forTest = new List <m_If>();
double best_scoreTRUE = 0; double best_scoreFALSE = 0; double best_scoreTotal = 0;
//-- add first feature----------
Feature local_first_S = new Feature();
if (methodSelection == TheEntropy.methodSelection_DCFS)
{
local_list_F.AddRange(DCFS_first_list_F);
local_first_S = this.DCFS_first_S.Copy();
}
else
{
local_list_F.AddRange(EnAlgo_first_list_F);
local_first_S = this.EnAlgo_first_S.Copy();
}
selected = local_first_S.name;
process02sub_optimizeFeature(local_rules_forTest, ref local_first_S, ref best_scoreTRUE, ref best_scoreFALSE, ref best_scoreTotal);
process02sub_checkGoOn(best_scoreTRUE, best_scoreFALSE, best_scoreTotal); //get initial score
local_list_S0.Add(local_first_S); //to add threshold
local_list_S.Add(selected);
local_rules_forTest.Add(TheMapData.convertFeature_to_If(local_first_S));
//-------------------
if (selection_autoNumber)
{
loopTime = local_list_F.Count();
}
for (int i = 2; i < loopTime + 2; i++)
{
if (local_list_F.Count == 0)
{
break;
} //out of data
if (methodSelection == TheEntropy.methodSelection_DCFS)
{
selected = DCFS_listDelta_getMaxValue();
if (path_prefix != "")
{
TheTool.exportCSV_orTXT(path_prefix + i + ".csv", DCFS_listDelta_getListString(true), false);
}
}
else
{
selected = TheEntropy.select1Candidate_entropy(methodSelection, dt_ready, local_list_F, local_list_S, list_entropy, list_MI);
if (path_prefix != "")
{
TheTool.exportCSV_orTXT(path_prefix + i + ".csv", TheTool.str_double_getListString(TheEntropy.temp_listEntropy), false);
}
}
Feature selected0 = TheData.getFeature(selected, dt_threshold_pose1, dt_threshold_pose2);
process02sub_optimizeFeature(local_rules_forTest, ref selected0, ref best_scoreTRUE, ref best_scoreFALSE, ref best_scoreTotal);
Boolean goOn = process02sub_checkGoOn(best_scoreTRUE, best_scoreFALSE, best_scoreTotal);
if (methodSelection == TheEntropy.methodSelection_DCFS)
{
DCFS_listDelta_update(selected);
}
if (selection_autoNumber == false || goOn)
{
local_list_F.Remove(selected);
local_list_S0.Add(selected0);
local_list_S.Add(selected);
local_rules_forTest.Add(TheMapData.convertFeature_to_If(selected0));
}
else
{
break;
}
}
return(local_list_S0);
}