private void save(string fullpath)
{
var model = this._model;
//提示是否保存光谱库，用于建模
this._model.FullPath = fullpath;
var finfo = new FileInfo(fullpath);
this._model.Name = finfo.Name.Replace(finfo.Extension, "");
if (MessageBox.Show("是否将建模所用的光谱库保存到模型中？ 备注：保存在模型中方可对模型进行修改。", "信息提示", MessageBoxButtons.YesNo, MessageBoxIcon.Question) == System.Windows.Forms.DialogResult.No)
{
model = Serialize.DeepClone <PLSModel>(this._model);
model.LibBase = null;
var lst = model.SubModels.Where(d => d.Trained == true).ToArray();
model.SubModels.Clear();
foreach (var m in lst)
{
model.SubModels.Add(m);
}
}
model.CreateTime = DateTime.Now;
model.Edited = false;
model.Trained = true;
if (model.Save())
{
MessageBox.Show("捆绑模型保存成功!", "信息提示", MessageBoxButtons.OK, MessageBoxIcon.Information);
this._model.Edited = false;
}
else
{
MessageBox.Show("捆绑模型保存失败!", "信息提示", MessageBoxButtons.OK, MessageBoxIcon.Information);
}
}
public void Train(SpecBase lib, bool needFilter = true)
{
//过滤掉性质有NaN的数据
lib.FilterNaN();
int[] idxs = lib.Specs.Select((d, idx) => new { s = d, idx = idx }).Where(d => d.s.Usage != UsageTypeEnum.Ignore).Select(d => d.idx).ToArray();
this._lib = Serialize.DeepClone <SpecBase>(lib.SubLib(idxs));
if (needFilter && this._filters != null)
{
this._lib.SetX(Preprocesser.Process(this._filters, this._lib), true);
}
//PCA分解
var handler = Tools.ModelHandler;
if (this._maxRank < 1)
{
this._maxRank = 10;
}
this._maxRank = Math.Min(this._maxRank, this._lib.X.Dimensions[1]);
var r = handler.IdentifyTrain(3, this._lib.X, this._maxRank);
this._p = (MWNumericArray)r[0];
this._w = (MWNumericArray)r[1];
this._t = (MWNumericArray)r[2];
this._trained = true;
}
private void btnModelExtent_Click(object sender, EventArgs e)
{
if (this._lib == null || this._baseModel == null)
{
return;
}
bool needcv = false;
if (MessageBox.Show("是否进行交互验证和外部验证？", "提示", MessageBoxButtons.YesNo, MessageBoxIcon.Question) == System.Windows.Forms.DialogResult.Yes)
{
needcv = true;
}
this._model = Serialize.DeepClone <IdentifyModel>(this._baseModel);
this._model.Name = this._lib.Name;
this.btnModelExtent.Text = "正在计算";
this.btnRebuild.Enabled = false;
this.btnModelExtent.Enabled = false;
this.btnLoadModel.Enabled = false;
this.btnLoadSpec.Enabled = false;
this.toolStripProgressBar1.Visible = true;
int num = int.Parse(this.textBox1.Text);
int numofId = int.Parse(this.txbnumofId.Text);
Action a = () =>
{
this._model.Train(this._lib);
var lib = this._model.SpecLib.Clone();
lib.Merger(this._baseModel.SpecLib);
if (needcv)
{
this._cvLst = this._model.CrossValidation(lib, false);
this._vLst = this._model.Validation(lib, false);
}
else
{
this._vLst = null;
this._cvLst = null;
}
this._model.Train(lib, false);
//显示
this.showdata(num, numofId);
if (this.InvokeRequired)
{
ThreadStart s = () =>
{
MessageBox.Show("建模完成");
};
this.Invoke(s);
}
else
{
MessageBox.Show("建模完成");
}
};
a.BeginInvoke(null, null);
}
public void LoadData(SpecBase filterlib, SpecBase baselib)
{
this._olib = Serialize.DeepClone <SpecBase>(filterlib);
this._nlib = Serialize.DeepClone <SpecBase>(baselib);
// this.renderGrid(UsageTypeEnum.Node);
if (cbxSamSelector.SelectedIndex == 0)
{
this.renderGrid(UsageTypeEnum.Node);
}
this.combRender();
}
public BindResult PredictForAPI(Spectrum spec, bool needFilter = true, int numOfId = 5, int topK = 1)
{
if (this.IdModels.Count > 0)
{
IdentifyResult iresult = null;
foreach (var i in this.IdModels)
{
iresult = CombineIdResult(iresult, i.Predict(spec, needFilter, numOfId, topK));
}
if (iresult != null && iresult.Items.Where(d => d.Result).Count() > 0)
{
iresult = IdentifyModel.GetPredictValue(iresult, iresult.Items.Length, numOfId);
return(new BindResult()
{
MethodType = PredictMethod.Identify,
Result = iresult
});
}
}
if (this.FitModels.Count > 0)
{
var fitmodel = Serialize.DeepClone <FittingModel>(this.FitModels.First());
for (int i = 1; i < this.FitModels.Count; i++)
{
fitmodel.SpecLib.Merger(this.FitModels[i].SpecLib);
}
var flst = new List <FittingResult>();//这里需要修改，将List<FittingResult>改为FittingResult
flst.Add(fitmodel.Predict(spec, needFilter, numOfId, topK));
if (flst.Where(d => d.Result).Count() > 0)
{
return new BindResult()
{
MethodType = PredictMethod.Fitting,
Result = flst.Where(d => d.Result).OrderByDescending(d => d.TQ).FirstOrDefault()
}
}
;
}
if (this.PLS != null)
{
var plsr = this.PLS.Predict(spec, needFilter, numOfId);
return(new BindResult()
{
MethodType = PredictMethod.PLSBind,
Result = plsr
});
}
return(new BindResult()
{
MethodType = PredictMethod.None
});
}
public void Train(SpecBase lib, bool needFilter = true)
{
//过滤掉性质有NaN的数据
lib.FilterNaN();
int[] idxs = lib.Specs.Select((d, idx) => new { s = d, idx = idx }).Where(d => d.s.Usage != UsageTypeEnum.Ignore).Select(d => d.idx).ToArray();
this._lib = Serialize.DeepClone <SpecBase>(lib.SubLib(idxs));
if (needFilter && this._filters != null)
{
this._lib.SetX(Preprocesser.Process(this._filters, this._lib), true);
}
this._trained = true;
}
private void btnSave_Click(object sender, EventArgs e)
{
try
{
this._model.MDMin = double.Parse(this.txbMash.Text);
this._model.SRMin = double.Parse(this.txbSR.Text);
this._model.Nonnegative = this.txbNotneg.Checked;
}
catch
{
MessageBox.Show("您输入的数字有误");
return;
}
SaveFileDialog mySaveFileDialog = new SaveFileDialog();
if (this._model.AnnType == PLSAnnEnum.None)
{
mySaveFileDialog.Filter = string.Format("{1} (*.{0})|*.{0}",
FileExtensionEnum.PLS1,
FileExtensionEnum.PLS1.GetDescription());
}
else
{
mySaveFileDialog.Filter = string.Format("{1} (*.{0})|*.{0}",
FileExtensionEnum.PLSANN,
FileExtensionEnum.PLSANN.GetDescription());
}
mySaveFileDialog.InitialDirectory = Busi.Common.Configuration.FolderMModelLib;
mySaveFileDialog.FileName = string.Format("{0}-{1}", this._model.LibBase.Name, this._model.Comp.Name);
if (mySaveFileDialog.ShowDialog() != System.Windows.Forms.DialogResult.OK)
{
return;
}
try
{
this._model.FullPath = mySaveFileDialog.FileName;
var model = Serialize.DeepClone <PLSSubModel>(this._model);
model.LibBase = null;
model.ParentModel = null;
model.Trained = true;
model.Save();
this.DialogResult = System.Windows.Forms.DialogResult.OK;
this.Close();
}
catch (Exception ex)
{
log.Error(ex);
MessageBox.Show("保存失败！");
}
}
void dataGridView2_CellEndEdit(object sender, DataGridViewCellEventArgs e)
{
if (e.ColumnIndex != 0)
{
return;
}
for (int i = 0; i < this.dataGridView2.Rows.Count; i++)
{
if (i != e.RowIndex)
{
this.dataGridView2.Rows[i].Cells[0].Value = false;
}
}
var cname = this.dataGridView2.Rows[e.RowIndex].Tag.ToString();
this._model.Comp = null;
if (this._model.Identify != null)
{
var c = this._model.Identify.Where(d => d.SpecLib.Components.Contains(cname)).Select(d => d.SpecLib.Components[cname]).FirstOrDefault();
if (c != null)
{
this._model.Comp = Serialize.DeepClone <Component>(c);
}
}
if (this._model.Comp == null && this._model.Fittings != null)
{
var c = this._model.Fittings.Where(d => d.SpecLib.Components.Contains(cname)).Select(d => d.SpecLib.Components[cname]).FirstOrDefault();
if (c != null)
{
this._model.Comp = Serialize.DeepClone <Component>(c);
}
}
if (this._model.Comp == null && this._model.PLS1 != null)
{
if (this._model.PLS1.Comp.Name == cname)
{
this._model.Comp = Serialize.DeepClone <Component>(this._model.PLS1.Comp);
}
}
if (this._model.Comp == null && this._model.PLSANN != null)
{
if (this._model.PLSANN.Comp.Name == cname)
{
this._model.Comp = Serialize.DeepClone <Component>(this._model.PLSANN.Comp);
}
}
}
private void btnSave_Click(object sender, EventArgs e)
{
var model = this._model;
if (model == null || model.LibBase == null || model.LibBase.Count == 0)
{
return;
}
if (String.IsNullOrWhiteSpace(this._model.FullPath))
{
SaveFileDialog mySaveFileDialog = new SaveFileDialog();
mySaveFileDialog.Filter = string.Format("{1} (*.{0})|*.{0}", FileExtensionEnum.FitLib, FileExtensionEnum.FitLib.GetDescription());
mySaveFileDialog.InitialDirectory = Busi.Common.Configuration.FolderMFit;
mySaveFileDialog.FileName = this._model.LibBase.Name;
if (mySaveFileDialog.ShowDialog() != DialogResult.OK)
{
return;
}
this._model.FullPath = mySaveFileDialog.FileName;
var finfo = new FileInfo(mySaveFileDialog.FileName);
this._model.Name = finfo.Name.Replace(finfo.Extension, "");
}
if (MessageBox.Show("是否将建模所用的光谱库保存到模型中？ 备注：保存在模型中方可对模型进行修改。", "信息提示", MessageBoxButtons.YesNo, MessageBoxIcon.Question) == System.Windows.Forms.DialogResult.No)
{
model = Serialize.DeepClone <FittingModel>(this._model);
model.LibBase = null;
}
model.Train(this._model.LibBase);
model.CreateTime = DateTime.Now;
model.Edited = false;
if (model.Save())
{
MessageBox.Show("拟合库保存成功!", "信息提示", MessageBoxButtons.OK, MessageBoxIcon.Information);
this._model.Edited = false;
}
else
{
MessageBox.Show("拟合库保存失败!", "信息提示", MessageBoxButtons.OK, MessageBoxIcon.Information);
}
this.setTitle();
}
/// <summary>
/// 添加新的预处理方法
/// </summary>
/// <param name="sender"></param>
/// <param name="e"></param>
void treeMethod_NodeMouseDoubleClick(object sender, TreeNodeMouseClickEventArgs e)
{
var node = e.Node as MyTreenode;
if (node == null)
{
return;
}
if (node.Preocesser.Filter.FType == FilterType.VarFilter)//如果是区间设置，需要先将之前的全部处理一次
{
Action a = () =>
{
this.compute();
var p = Serialize.DeepClone <Preprocesser>(node.Preocesser);
p.SpecsInput = this._lastRowSpec;
var dlg = new VarRegionSet();
dlg.Width = (int)(this.TopLevelControl.Width * 0.85);
dlg.Height = (int)(this.TopLevelControl.Height * 0.9);
NIR.Component c = null;
if (this.GetComponent != null)
{
c = this.GetComponent();
}
if (c == null)
{
c = p.SpecsInput.Components.FirstOrDefault();
}
if (dlg.ShowDialog(p, c) == DialogResult.OK)
{
this.addRow(p);
}
};
this.BeginInvoke(a);
}
else
{
this.addRow(Serialize.DeepClone <Preprocesser>(node.Preocesser));
}
}
private void btnRebuild_Click(object sender, EventArgs e)
{
if (this._lib == null || this._baseModel == null)
{
return;
}
bool needcv = false;
if (MessageBox.Show("是否进行交互验证和外部验证？", "提示", MessageBoxButtons.YesNo, MessageBoxIcon.Question) == System.Windows.Forms.DialogResult.Yes)
{
needcv = true;
}
this._model = Serialize.DeepClone <PLSSubModel>(this._baseModel);
this.btnRebuild.Text = "正在计算";
this.btnRebuild.Enabled = false;
this.btnLoadModel.Enabled = false;
this.btnLoadSpec.Enabled = false;
this.toolStripProgressBar1.Visible = true;
Action a = () =>
{
if (needcv)
{
this._cvLst = this._model.CrossValidation(this._lib);
this._vLst = this._model.Validation(this._lib);
}
else
{
this._cvLst = null;
this._vLst = null;
}
this._model.Train(this._lib);
//显示
this.showdata();
};
a.BeginInvoke(null, null);
}
private void train()
{
try
{
for (int i = 0; i < this._Nodes.Count; i++)
{
var pls = this._Nodes[i].PLS.Model;
var needtrain = (bool)this.dataGridView1.Rows[i].Cells[3].Value;
var needOuter = (bool)this.dataGridView1.Rows[i].Cells[4].Value;
if (pls.Comp == this._model.Comp)
{
continue;
}
if (needtrain)
{
// Pharse 1 设置方法和参数
// 预处理
pls.Filters = new List <IFilter>();
foreach (var f in _model.Filters)
{
pls.Filters.Add(Serialize.DeepClone <IFilter>(f));
}
pls.MaxFactor = this._model.MaxFactor;
pls.Factor = this._model.Factor;
pls.ANNAgrus = this._model.ANNAgrus;
pls.ANNModel = this._model.ANNModel;
pls.AnnType = this._model.AnnType;
pls.FANNModels = this._model.FANNModels;
pls.Method = this._model.Method;
pls.SRMin = this._model.SRMin;
pls.Nonnegative = this._model.Nonnegative;
if (needOuter)
{
pls.OutlierNames = Serialize.DeepClone <List <string> >(this._model.OutlierNames);
}
// Pharse 2 交互验证
this.c(i, "正在交互验证");
var dt = DateTime.Now;
this._Nodes[i].PLS.CVResult = pls.CrossValidation(pls.LibBase);
// Pharse 3 外部验证
this.c(i, "正在外部验证");
//pls.Train(pls.LibBase);
//var lst = pls.LibBase.Where(d => d.Usage == UsageTypeEnum.Validate);
var span1 = (DateTime.Now - dt).TotalMilliseconds;
//再外部验证
dt = DateTime.Now;
this._Nodes[i].PLS.VResult = pls.Validation(pls.LibBase);
var span2 = (DateTime.Now - dt).TotalMilliseconds;
log.DebugFormat("{0}个校正集，交互验证花费{1}ms，{2}个验证集，外部验证花费{3}ms", pls.LibBase.Specs.Where(d => d.Usage == UsageTypeEnum.Calibrate).Count(),
span1,
pls.LibBase.Specs.Where(d => d.Usage == UsageTypeEnum.Validate).Count(),
span2);
pls.MDMin = pls.Mdt[pls.MaxFactor - 1];
pls.NNDMin = pls.NNdt[pls.MaxFactor - 1];
// Pharse 4
this.c(i, "完成");
this._Nodes[i].PLS.ActiveStep = 4;
var parent = this._Nodes[i].TreeView;
if (parent.InvokeRequired)
{
ThreadStart s = () =>
{
this._Nodes[i].ShowText();
};
parent.Invoke(s);
}
else
{
this._Nodes[i].ShowText();
}
//使用完后立即回收lib
if (pls.Lib != null)
{
pls.Lib.Dispose();
}
}
}
}
catch (OutOfMemoryException ex)
{
log.Error(ex);
MessageBox.Show("对不起，性质数量太多，内存溢出了，请关闭本窗口，保存已建好性质的子模型后，重启本软件，谢谢！");
}
finally
{
if (this.button1.InvokeRequired)
{
ThreadStart s = () => { this.button1.Enabled = true; };
this.button1.Invoke(s);
}
else
{
this.button1.Enabled = true;
}
if (this.button2.InvokeRequired)
{
ThreadStart s = () => { this.button2.Enabled = true; };
this.button2.Invoke(s);
}
else
{
this.button1.Enabled = true;
}
if (this.progressBar1.InvokeRequired)
{
ThreadStart s = () => { this.progressBar1.Visible = false; };
this.progressBar1.Invoke(s);
}
else
{
this.progressBar1.Visible = false;
}
this._busying = false;
}
}
public Component Clone()
{
return(Serialize.DeepClone <Component>(this));
}
public PLS1Result[] CrossValidation(SpecBase libb, bool needFilter = true, int numOfId = 5)
{
var lib = libb.Clone();
//过滤掉性质有NaN的数据
lib.FilterNaN(this._comp);
foreach (var s in lib.Specs)
{
if (this._OutlierNames.Contains(s.Name))
{
s.Usage = UsageTypeEnum.Ignore;
}
}
//lib.Specs = lib.Specs.Where(d => !this._OutlierNames.Contains(d.Name)).ToList();
if (!this._trained)
{
this.Train(lib, needFilter);
}
var clib = lib.SubLib(UsageTypeEnum.Calibrate);
if (clib.Count == 0)
{
return(null);
}
if (needFilter && this._filters != null)
{
clib.X = Preprocesser.Process(this._filters, clib.X);
}
var handler = Tools.ModelHandler;
MWNumericArray Ylast, SR, MD, nd;
if (this._annType != PLSAnnEnum.None)
{
var model = Serialize.DeepClone <PLSSubModel>(this);
model.Train(lib, true);
var pls = handler.PLS1Predictor(5, clib.X, model.Scores, model.Loads, model.Weights, model.Bias, model.Score_Length, model.CenterSpecData, model.CenterCompValue, (int)this._method);
var toolHandler = Tools.ToolHandler;
Ylast = (MWNumericArray)pls[0];
SR = (MWNumericArray)pls[1];
MD = (MWNumericArray)pls[2];
nd = (MWNumericArray)pls[3];
var mscores = (MWNumericArray)toolHandler.Centring(1, pls[4], model.ScoresMean)[0];
double[] annrsult;
var annp = handler.annp(1,
mscores,
model.ANNModel.w1,
model.ANNModel.b1,
model.ANNModel.w2,
model.ANNModel.b2,
model.ANNAgrus.F1.GetDescription(),
model.ANNAgrus.F2.GetDescription())[0];
annrsult = (double[])((MWNumericArray)annp).ToVector(MWArrayComponent.Real);
for (int row = 0; row < annrsult.Length; row++)
{
Ylast[row + 1, this._factor] = annrsult[row] + model.CenterCompValue;
}
}
else
{
var pls = handler.PLS1CrossValidation(4, clib.X, clib.GetY(this._comp, true), this._maxFactor, (int)this._method);
Ylast = (MWNumericArray)pls[0];
SR = (MWNumericArray)pls[1];
MD = (MWNumericArray)pls[2];
nd = (MWNumericArray)pls[3];
}
var items = new List <PLS1Result>();
for (int i = 0; i < clib.Count; i++)
{
var c = this._comp.Clone();
var s = clib[i];
if (s.Components != null && s.Components.Contains(c.Name))
{
c.ActualValue = s.Components[c.Name].ActualValue;
}
var r = new PLS1Result
{
MDMin = this._MDMin,
NDMin = this._NNDMin,
SRMin = this._SRMin,
Comp = c,
Spec = s,
YLast = (double[])Tools.SelectRow(Ylast, i + 1).ToVector(MWArrayComponent.Real),
SR = (double[])Tools.SelectRow(SR, i + 1).ToVector(MWArrayComponent.Real),
MahDist = (double[])Tools.SelectRow(MD, i + 1).ToVector(MWArrayComponent.Real),
ND = (double[])Tools.SelectRow(nd, i + 1).ToVector(MWArrayComponent.Real),
Factor = this.Factor
};
if (!double.IsNaN(r.YLast[0]))
{
items.Add(r);
}
}
lib.Dispose();
clib.Dispose();
GC.Collect();
return(items.ToArray());
}
public static IdentifyResult GetPredictValue(IdentifyResult input, int num, int numOfId)
{
if (input == null || input.Items == null)
{
throw new ArgumentNullException("");
}
var r = Serialize.DeepClone <IdentifyResult>(input);
//重新复制原始光谱和wind
for (int i = 0; i < r.Items.Length; i++)
{
r.Items[i].SpecOriginal = input.Items[i].SpecOriginal;
r.Items[i].Wind = input.Items[i].Wind;
}
numOfId = numOfId < 1 ? 5 : numOfId;
var predicts = r.Items.Where(d => d.Result).OrderByDescending(d => d.Result).ThenByDescending(d => d.TQ).Select(d => d.Spec.Components).Take(num);
// var count = predicts.Count();
if (r.Components == null)
{
r.Components = r.Items.First().Spec.Components.Clone();
}
int k = 0;
int[] couter = new int[r.Components.Count];
int numCounter = 0;
foreach (var p in predicts)
{
if (k == 0)
{
foreach (var c in r.Components)
{
c.PredictedValue = 0;
}
k++;
}
for (int i = 0; i < r.Components.Count; i++)
{
if (!double.IsNaN(p[i].ActualValue))
{
r.Components[i].PredictedValue += p[i].ActualValue;
couter[i]++;
}
}
numCounter++;
if (numCounter >= numOfId)
{
break;
}
}
for (int i = 0; i < r.Components.Count; i++)
{
r.Components[i].PredictedValue = r.Components[i].PredictedValue / couter[i];
r.Components[i].Error = r.Components[i].PredictedValue - r.Components[i].ActualValue;
}
r.Spec = input.Spec;
r.Items = r.Items.OrderByDescending(d => d.Result).ThenByDescending(d => d.TQ).ToArray();
return(r);
}
private void btnCompute_Click(object sender, EventArgs e)
{
if (spec1 == null || spec2 == null)
{
MessageBox.Show(string.Format("第{0}条光谱为空，请先选择。",
spec1 == null ? 1 : 2), "信息提示");
return;
}
int mwin = 0;
if (!int.TryParse(txbmwin.Text, out mwin))
{
MessageBox.Show("移动窗口大小必须为整数", "信息提示");
return;
}
IList <IFilter> filters = new List <IFilter>();
if (this._model != null)
{
if (this._model is IdentifyModel)
{
filters = ((IdentifyModel)this._model).Filters;
}
else if (this._model is FittingModel)
{
filters = ((FittingModel)this._model).Filters;
}
}
var s1 = spec1.Clone();
var s2 = spec2.Clone();
//对光谱预处理
if (filters != null)
{
var ya = new MWNumericArray(s1.Data.Lenght, 1, s1.Data.Y);
var yb = new MWNumericArray(s2.Data.Lenght, 1, s2.Data.Y);
bool splitadd = false;
var y1 = (MWNumericArray)ya.Clone();
var y2 = (MWNumericArray)yb.Clone();
var x1 = Serialize.DeepClone <double[]>(s1.Data.X);
var x2 = Serialize.DeepClone <double[]>(s2.Data.X);
var y1lst = new List <MWNumericArray>();
var x1lst = new List <double[]>();
var y2lst = new List <MWNumericArray>();
var x2lst = new List <double[]>();
foreach (var tf in filters)
{
var f = RIPP.Lib.Serialize.DeepClone <RIPP.NIR.Data.Filter.IFilter>(tf);
if (f is Spliter)
{
if (splitadd)
{
x1lst.Add(x1);
y1lst.Add(y1);
x2lst.Add(x2);
y2lst.Add(y2);
}
splitadd = false;
y1 = (MWNumericArray)ya.Clone();
y2 = (MWNumericArray)yb.Clone();
x1 = Serialize.DeepClone <double[]>(s1.Data.X);
x2 = Serialize.DeepClone <double[]>(s2.Data.X);
}
else
{
splitadd = true;
}
y1 = f.Process(y1);
y2 = f.Process(y2);
if (f.FType == FilterType.VarFilter)
{
x1 = f.VarProcess(x1);
x2 = f.VarProcess(x2);
}
}
if (splitadd)
{
x1lst.Add(x1);
y1lst.Add(y1);
x2lst.Add(x2);
y2lst.Add(y2);
}
//合并
if (x1lst.Count > 0)
{
s1.Data.X = x1lst[0];
s1.Data.Y = (double[])y1lst[0].ToVector(MWArrayComponent.Real);
s2.Data.X = x2lst[0];
s2.Data.Y = (double[])y2lst[0].ToVector(MWArrayComponent.Real);
for (int i = 1; i < x1lst.Count; i++)
{
s1.Data.X = RIPP.NIR.Data.Tools.InsertColumn(s1.Data.X, x1lst[i], s1.Data.X.Length + 1);
s1.Data.Y = RIPP.NIR.Data.Tools.InsertColumn(s1.Data.Y, (double[])y1lst[i].ToVector(MWArrayComponent.Real), s1.Data.Y.Length + 1);
s2.Data.X = RIPP.NIR.Data.Tools.InsertColumn(s2.Data.X, x2lst[i], s2.Data.X.Length + 1);
s2.Data.Y = RIPP.NIR.Data.Tools.InsertColumn(s2.Data.Y, (double[])y2lst[i].ToVector(MWArrayComponent.Real), s2.Data.Y.Length + 1);
}
}
else
{
s1.Data.X = x1;
s1.Data.Y = (double[])y1.ToVector(MWArrayComponent.Real);
s2.Data.X = x2;
s2.Data.Y = (double[])y2.ToVector(MWArrayComponent.Real);
}
//s1.Data.Y = (double[])y1.ToVector(MWArrayComponent.Real);
//s2.Data.Y = (double[])y2.ToVector(MWArrayComponent.Real);
}
if (this._model is FittingModel)
{
var f = ((FittingModel)this._model).IdRegion;
s1.Data.X = f.VarProcess(s1.Data.X);
s2.Data.X = f.VarProcess(s2.Data.X);
s1.Data.Y = f.VarProcess(s1.Data.Y);
s2.Data.Y = f.VarProcess(s2.Data.Y);
}
//绘制
var lst = new List <Spectrum>();
lst.Add(s1);
lst.Add(s2);
this.specGraph3.DrawSpec(lst);
this.specGraph3.SetTitle("预处理后光谱");
double[] tSQ;
double tTQ;
RIPP.NIR.Data.Tools.MWCorr(s1.Data.Y, s2.Data.Y, mwin, out tTQ, out tSQ);
var spec = RIPP.Lib.Serialize.DeepClone <Spectrum>(s1);
spec.Data.Y = tSQ;
spec.Name = "SQ";
spec.Color = Color.Blue;
this.specGraph2.DrawSpec(new List <Spectrum>()
{
spec
});
this.txbTQ.Text = tTQ.ToString("f3");
this.txbSQ.Text = tSQ.Min().ToString("f3");
}
public SpecBase Clone()
{
return(Serialize.DeepClone <SpecBase>(this));
}
public Spectrum Clone()
{
return(Serialize.DeepClone <Spectrum>(this));
}
private void btnLoadModel_Click(object sender, EventArgs e)
{
OpenFileDialog myOpenFileDialog = new OpenFileDialog();
myOpenFileDialog.Filter = string.Format("{0} (*.{1})|*.{1}", FileExtensionEnum.PLSBind.GetDescription(), FileExtensionEnum.PLSBind);
myOpenFileDialog.InitialDirectory = Busi.Common.Configuration.FolderBlendMod;
if (myOpenFileDialog.ShowDialog() == DialogResult.OK)
{
var m = BindModel.ReadModel <PLSModel>(myOpenFileDialog.FileName);
if (m == null || lib == null)
{
return;
}
if (m.SubModels.Count != this.lib.Components.Count)
{
MessageBox.Show(string.Format("参与混兑计算的光谱数量为{0}，加载的捆绑模型子模型数据为{1}，不匹配。", this.lib.Components.Count, m.SubModels.Count));
return;
}
this.toolStrip2.Enabled = false;
this.btnLoadModel.Text = "正在计算";
this.progressBar1.Visible = true;
Action a = () =>
{
this.model = Serialize.DeepClone <PLSModel>(m);
this.model.MixSpecs = new List <Spectrum>();
if (s1 != null)
{
this.model.MixSpecs.Add(s1);
}
if (s2 != null)
{
this.model.MixSpecs.Add(s2);
}
if (s3 != null)
{
this.model.MixSpecs.Add(s3);
}
int i = 0;
foreach (var sm in this.model.SubModels)
{
sm.Comp = this.lib.Components[i];
sm.Train(this.lib);
var cvlst = sm.CrossValidation(this.lib);
if (this.specGridView1.InvokeRequired)
{
ThreadStart s = () => { this.add(cvlst, i); };
this.specGridView1.Invoke(s);
}
else
{
this.add(cvlst, i);
}
i++;
}
if (this.toolStrip2.InvokeRequired)
{
ThreadStart s = () =>
{
this.progressBar1.Visible = false;
this.btnLoadModel.Text = "应用模型";
this.toolStrip2.Enabled = true;
};
this.toolStrip2.Invoke(s);
}
else
{
this.progressBar1.Visible = false;
this.btnLoadModel.Text = "应用模型";
this.toolStrip2.Enabled = true;
}
};
a.BeginInvoke(null, null);
}
}
public void Train(Matrix x, Vector y)
{
//检查输入
if (x == null || y == null)
{
throw new ArgumentNullException("");
}
if (x.ColumnCount != y.Count)
{
throw new ArgumentException("x,y长度不一致");
}
if (this._MaxFactor <= 0)
{
throw new ArgumentOutOfRangeException(string.Format("Factor未设置或者为负，Factor={0}", this._MaxFactor));
}
if (this._MaxFactor > x.RowCount)
{
throw new ArgumentOutOfRangeException(string.Format("主成分数不能大于样品数,Factor={0}，Num ={1}", this._MaxFactor, x.RowCount));
}
if (this._MaxFactor > x.ColumnCount)
{
throw new ArgumentOutOfRangeException(string.Format("主成分数不能大于波长点数,Factor={0}，Num ={1}", this._MaxFactor, x.ColumnCount));
}
this._x = Serialize.DeepClone <Matrix>(x);
this._y = Serialize.DeepClone <Vector>(y);
int sampleCount = this._x.ColumnCount;
int pointCount = this._x.RowCount;
this._centerSpecData = new DenseVector(pointCount);
for (int i = 0; i < pointCount; i++)
{
var r = this._x.Row(i);
this._centerSpecData[i] = r.Mean();
this._x.SetRow(i, r.Subtract(this._centerSpecData[i]));
}
this._centerCompValue = y.Mean();
y = (Vector)y.Subtract(this._centerCompValue);
this._x = (Matrix)this._x.Transpose();
this._Weights = new DenseMatrix(pointCount, this._MaxFactor);
this._Loads = new DenseMatrix(pointCount, this._MaxFactor);
this._Scores = new DenseMatrix(sampleCount, this._MaxFactor);
this._Bias = new DenseVector(this._MaxFactor);
this._Score_Length = new DenseVector(this._MaxFactor);
for (int f = 0; f < this._MaxFactor; f++)
{
var xt = this._x.Transpose();
var w = xt * this._y;
var s = this._x * w;
this._Score_Length[f] = Math.Sqrt(s.DotProduct(s));
w = w * this._Score_Length[f];
s = s / this._Score_Length[f];
var l = xt * s;
this._Bias[f] = s.DotProduct(y);
this._Weights.SetColumn(f, w);
this._Scores.SetColumn(f, s);
this._Loads.SetColumn(f, l);
this._x = (Matrix)(this._x - s.OuterProduct(l));
this._y = (Vector)(y - s * this._Bias[f]);
}
this._isTrained = true;
}
public List <PropertyTable> GetData(Specs s, BindModel predictor)
{
using (var db = new NIRCeneterEntities())
{
var r = s.OilData;
if (r != null)
{
return(r);
}
if (s == null || s.ResultObj == null)
{
return(null);
}
var pr = s.ResultObj;
if (s.ResultObj.MethodType == PredictMethod.Integrate && predictor != null)
{
var papi = predictor.PredictForAPI(s.Spec, true);
if (papi.MethodType == PredictMethod.Fitting || papi.MethodType == PredictMethod.Identify)
{
pr = papi;
}
}
r = Serialize.DeepClone <List <PropertyTable> >(this._initP);
OilInfoBEntity oil = null;
switch (pr.MethodType)
{
case NIR.Models.PredictMethod.Identify:
oil = getByName(s, pr, ref r);
break;
case NIR.Models.PredictMethod.Fitting:
oil = getByRate(s, pr, ref r);
break;
case NIR.Models.PredictMethod.PLSBind:
oil = getByProperties(s, pr, ref r);
break;
default:
break;
}
GetNIRData(s, oil, ref r);
if (oil != null)
{
r = r.OrderBy(d => (int)d.Table).ToList();
db.OilData.AddObject(new OilData()
{
SID = s.ID,
Data = Serialize.ObjectToByte(r)
});
db.SaveChanges();
}
else
{
foreach (var t in r)
{
foreach (var dd in t.Datas)
{
if (dd.Value == 0)
{
dd.Value = double.NaN;
}
}
}
}
return(r);
}
}
public bool APIGetData(BindModel predictor, ref Specs s)
{
log.Info("db before");
using (var db = new NIRCeneterEntities())
{
var r = s.OilData;
if (r != null)
{
return(true);
}
log.Info("db after");
if (s == null || s.ResultObj == null)
{
return(false);
}
var pr = s.ResultObj;
if (s.ResultObj.MethodType == PredictMethod.Integrate && predictor != null)
{
var papi = predictor.PredictForAPI(s.Spec, true);
if (papi.MethodType == PredictMethod.Fitting || papi.MethodType == PredictMethod.Identify)
{
pr = papi;
}
}
r = Serialize.DeepClone <List <PropertyTable> >(this._initP);
OilInfoBEntity oil = null;
switch (pr.MethodType)
{
case NIR.Models.PredictMethod.Identify:
oil = getByName(s, pr, ref r);
break;
case NIR.Models.PredictMethod.Fitting:
oil = getByRate(s, pr, ref r);
break;
case PredictMethod.Integrate:
case NIR.Models.PredictMethod.PLSBind:
oil = getByProperties(s, pr, ref r);
break;
default:
break;
}
GetNIRData(s, oil, ref r);
if (oil != null)
{
r = r.OrderBy(d => (int)d.Table).ToList();
db.OilData.AddObject(new OilData()
{
SID = s.ID,
Data = Serialize.ObjectToByte(r)
});
db.SaveChanges();
s.OilData = r;
return(true);
}
else
{
s.OilData = null;
}
return(false);
}
}
// [TestMethod]
public void fiiting()
{
var specbase = new SpecBase(@"F:\3506\15chemometrics\RIPP_DEMO\algorithm\原油测试数据\crude.Lib");
var model = new FittingModel()
{
Wind = 11, MinSQ = 0.98, TQ = 0.998
};
// var model = new IdentifyModel() { Wind = 11, MinSQ = 0.98, TQ = 0.998 };
model.Filters = new List <IFilter>();
model.Filters.Add(new Sgdiff(21, 2, 2));
model.Filters.Add(new SavitzkyGolay(5));
var xidx = new List <RegionPoint>();
xidx.Add(new RegionPoint(4002, 4702));
xidx.Add(new RegionPoint(5302, 6102));
var varregion = new VarRegionManu();
var argu = varregion.Argus;
argu["XaxisRegion"].Value = xidx;
argu["Xaxis"].Value = specbase.First().Data.X;
varregion.Argus = argu;
model.Filters.Add(varregion);
model.Filters.Add(new NormPathLength());
model.FiltersForIdentify = Serialize.DeepClone <IList <IFilter> >(model.Filters);
var vspecbase = specbase.Clone();
vspecbase.Clear();
foreach (var s in specbase)
{
if (s.Usage == UsageTypeEnum.Calibrate)
{
vspecbase.Add(s.Clone());
}
}
var dd = RIPP.NIR.Data.Preprocesser.Process(model.Filters, vspecbase);
var writer = new MatlabMatrixWriter(@"F:\3506\15chemometrics\RIPP_DEMO\src\RIPP\testdata\fit\filterd.mat");
writer.WriteMatrix <double>(dd.GetX(), "filterd");
writer.Close();
var cv = new CrossValidation <FittingResult>(model);
var lst = cv.CV(specbase);
var allTQ = new DenseVector(specbase.Where(d => d.Usage == UsageTypeEnum.Calibrate).Count());
var allSQ = new DenseVector(specbase.Where(d => d.Usage == UsageTypeEnum.Calibrate).Count());
var allRate = new DenseMatrix(5, specbase.Where(d => d.Usage == UsageTypeEnum.Calibrate).Count());
for (int i = 0; i < lst.Count; i++)
{
allTQ[i] = lst[i].TQ;
allSQ[i] = lst[i].SQ.Min();
var ddd = lst[i].Specs.Select(d => d.Rate);
for (int k = 0; k < 5; k++)
{
if (k >= ddd.Count())
{
break;
}
allRate[k, i] = ddd.ElementAt(k);
}
}
writer = new MatlabMatrixWriter(@"F:\3506\15chemometrics\RIPP_DEMO\src\RIPP\testdata\fit\tq.mat");
writer.WriteMatrix <double>(allTQ.ToColumnMatrix(), "TQ");
writer.Close();
writer = new MatlabMatrixWriter(@"F:\3506\15chemometrics\RIPP_DEMO\src\RIPP\testdata\fit\sq.mat");
writer.WriteMatrix <double>(allSQ.ToColumnMatrix(), "SQ");
writer.Close();
writer = new MatlabMatrixWriter(@"F:\3506\15chemometrics\RIPP_DEMO\src\RIPP\testdata\fit\result.mat");
writer.WriteMatrix <double>(allRate, "allRate");
writer.Close();
}
public List <IntegrateResultItem> Predict(Spectrum spec, bool needFilter = true, int numOfId = 5, int topK = 1)
{
foreach (var c in this._comps)
{
c.PredictedValue = 0;
}
ComponentList idresult = this._comps.Clone();
ComponentList fitresult = this._comps.Clone();
ComponentList pls1result = this._comps.Clone();
ComponentList plsannresult = this._comps.Clone();
IdentifyItem[][] idobj = new IdentifyItem[this._comps.Count][];
FittingResult[] fitobj = new FittingResult[this._comps.Count];
PLS1Result[] pls1obj = new PLS1Result[this._comps.Count];
PLS1Result[] annobj = new PLS1Result[this._comps.Count];
int[] groupidid = new int[this._comps.Count];
int[] groupfitid = new int[this._comps.Count];
double[] idTQ = new double[this._comps.Count];
double[] fitTQ = new double[this._comps.Count];
double[] idSQ = new double[this._comps.Count];
double[] fitSQ = new double[this._comps.Count];
bool[] okid = new bool[this._comps.Count];
bool[] okfit = new bool[this._comps.Count];
bool[] okpls1 = new bool[this._comps.Count];
bool[] okann = new bool[this._comps.Count];
bool[] innerPLs1 = new bool[this._comps.Count];
bool[] innerAnn = new bool[this._comps.Count];
double[] minIdTQ = new double[this._comps.Count];
double[] minFitTQ = new double[this._comps.Count];
double[] minIdSQ = new double[this._comps.Count];
double[] minFitSQ = new double[this._comps.Count];
DateTime dt = DateTime.Now;
if (this.IdModels != null && this.IdModels.Count > 0)
{
var ilst = new List <IdentifyResult>();
foreach (var i in this.IdModels)
{
ilst.Add(i.Predict(spec, needFilter, numOfId, topK));
}
var group = this._idSelectMatrix.Select((d, i) => new { idx = i, d = d }).GroupBy(d => d.d);
foreach (var g in group)
{
var a = new List <IdentifyItem>();
for (int i = 0; i < this.IdModels.Count; i++)
{
var d = (int)Math.Pow(2, i);
if ((d & g.Key) == d)
{
a.AddRange(ilst[i].Items);
}
}
//找识别出来的
var clst = a.OrderByDescending(d => d.Result).ThenByDescending(d => d.TQ);
//取识别结果的明细
var idgobj = clst.Take(10).ToArray();//需要修改
if (clst.Count() > 0)
{
var idtrue = clst.Where(d => d.Result);
foreach (var i in g)
{
minIdTQ[i.idx] = clst.First().TQ;
minIdSQ[i.idx] = clst.First().SQ;
var allc = idtrue.Select(d => d.Spec.Components).Where(d => d.Contains(idresult[i.idx].Name)).Select(d => d[idresult[i.idx].Name].ActualValue).Take(numOfId);
if (allc.Count() > 0)
{
okid[i.idx] = true;
idresult[i.idx].PredictedValue = allc.Sum() / allc.Count();
}
else
{
var cs = clst.First().Spec.Components;
idresult[i.idx].PredictedValue = cs.Contains(idresult[i.idx].Name) ? cs[idresult[i.idx].Name].ActualValue : double.NaN;
}
idobj[i.idx] = idgobj;
groupidid[i.idx] = g.Key;
idTQ[i.idx] = this.IdModels.First().TQ;
idSQ[i.idx] = this.IdModels.First().MinSQ;
}
}
}
}
log.InfoFormat("识别库花费{0}s，共{1}个", (DateTime.Now - dt).TotalSeconds, this.IdModels.Count);
dt = DateTime.Now;
if (this.FitModels != null && this.FitModels.Count > 0)
{
var group = this._fitSelectMatrix.Select((d, i) => new { idx = i, d = d }).GroupBy(d => d.d);
var ilst = new List <FittingResult>();
int gidx = -1;
foreach (var g in group)
{
gidx++;
if (this._cacheFitModel == null)
{
this._cacheFitModel = new List <FittingModel>();
}
if (this._cacheFitModel.Count < gidx + 1)
{
var models = new List <FittingModel>();
for (int i = 0; i < this.FitModels.Count; i++)
{
var d = (int)Math.Pow(2, i);
if ((d & g.Key) == d)
{
models.Add(this.FitModels[i]);
}
}
if (models.Count == 0)
{
continue;
}
var idxmax = models.Select(d => d.SpecLib.Count).Select((d, i) => new { idx = i, d = d }).OrderByDescending(d => d.d).First().idx;
var m = Serialize.DeepClone <FittingModel>(models[idxmax]);
//合并
DateTime dtn = DateTime.Now;
for (int i = 0; i < models.Count; i++)
{
if (i != idxmax)
{
m.SpecLib.Merger(models[i].SpecLib);
}
}
log.InfoFormat("merger spend {0}s", (DateTime.Now - dtn).TotalSeconds);
dtn = DateTime.Now;
this._cacheFitModel.Add(m);
}
var dddt = DateTime.Now;
var fresult = this._cacheFitModel[gidx].Predict(spec, needFilter, numOfId, topK);
log.InfoFormat("fit predict spend {0}s", (DateTime.Now - dddt).TotalSeconds);
foreach (var i in g)
{
minFitTQ[i.idx] = fresult.TQ;
minFitSQ[i.idx] = fresult.SQ;
okfit[i.idx] = fresult.Result;
var cs = fresult.FitSpec.Components;
fitresult[i.idx].PredictedValue = cs.Contains(fitresult[i.idx].Name) ? cs[fitresult[i.idx].Name].PredictedValue : double.NaN;
fitobj[i.idx] = fresult;
groupfitid[i.idx] = g.Key;
fitTQ[i.idx] = this._cacheFitModel[gidx].TQ;
fitSQ[i.idx] = this._cacheFitModel[gidx].MinSQ;
}
}
}
log.InfoFormat("拟合库花费{0}s，共{1}个", (DateTime.Now - dt).TotalSeconds, this.FitModels.Count);
dt = DateTime.Now;
if (this.Pls1Models != null && this.Pls1Models.Count > 0)
{
foreach (var m in this.Pls1Models)
{
var idx = pls1result.GetIndex(m.Comp.Name);
if (idx >= 0)
{
var pre = m.Predict(spec, needFilter);
pls1result[idx].PredictedValue = pre.Comp.PredictedValue;
okpls1[idx] = pre.Comp.State != ComponentStatu.Red;
innerPLs1[idx] = pre.Comp.State == ComponentStatu.Pass;
pls1obj[idx] = pre;
}
}
}
log.InfoFormat("PLS1花费{0}s，共{1}个", (DateTime.Now - dt).TotalSeconds, this.Pls1Models.Count);
dt = DateTime.Now;
if (this.PlsANNModels != null && this.PlsANNModels.Count > 0)
{
foreach (var m in this.PlsANNModels)
{
var idx = plsannresult.GetIndex(m.Comp.Name);
if (idx >= 0)
{
var pre = m.Predict(spec, needFilter);
plsannresult[idx].PredictedValue = pre.Comp.PredictedValue;
okann[idx] = pre.Comp.State != ComponentStatu.Red;
innerAnn[idx] = pre.Comp.State == ComponentStatu.Pass;
annobj[idx] = pre;
}
}
}
log.InfoFormat("PLSANN花费{0}s，共{1}个", (DateTime.Now - dt).TotalSeconds, this.PlsANNModels.Count);
//开始算系数
var result = new List <IntegrateResultItem>();
for (int i = 0; i < this._comps.Count; i++)
{
var c = this._comps[i];
var r = new IntegrateResultItem()
{
Comp = c.Clone(),
MinFitTQ = minFitTQ[i] == 0 ? double.NaN : minFitTQ[i],
MinIdTQ = minIdTQ[i] == 0 ? double.NaN : minIdTQ[i],
MinFitSQ = minFitSQ[i] == 0 ? double.NaN : minFitSQ[i],
MinIdSQ = minIdSQ[i] == 0 ? double.NaN : minIdSQ[i]
};
r.PrimalWID = this._weights[i, 0];
r.PrimalWFit = this._weights[i, 1];
r.PrimalWPLS1 = this._weights[i, 2];
r.PrimalWANN = this._weights[i, 3];
//取系数
r.IdWeight = this._weights[i, 0] > 0 ? this._weights[i, 0] : double.NaN;
r.FitWeight = this._weights[i, 1] > 0 ? this._weights[i, 1] : double.NaN;
r.Pls1Weight = this._weights[i, 2] > 0 ? this._weights[i, 2] : double.NaN;
r.ANNWeight = this._weights[i, 3] > 0 ? this._weights[i, 3] : double.NaN;
//取值
r.IdOk = okid[i];
r.FitOk = okfit[i];
r.PLS1Ok = okpls1[i];
r.ANNOk = okann[i];
r.InnerANN = innerAnn[i];
r.InnerPLS = innerPLs1[i];
r.IdValue = idresult[c.Name].PredictedValue;
r.FitValue = fitresult[c.Name].PredictedValue;
r.Pls1Value = okpls1[i] ? pls1result[c.Name].PredictedValue : double.NaN;
r.ANNValue = okann[i] ? plsannresult[c.Name].PredictedValue : double.NaN;
r.IdResult = idobj[i];
r.FitResult = fitobj[i];
r.Pls1Result = pls1obj[i];
r.AnnResult = annobj[i];
r.GroupIDID = groupidid[i];
r.GroupFitID = groupfitid[i];
r.IDSQ = idSQ[i];
r.IDTQ = idTQ[i];
r.FitSQ = fitSQ[i];
r.FitTQ = fitTQ[i];
//计算最终结果
bool failid = r.IdResult != null && okid[i] == false;
bool failfit = r.FitResult != null && okfit[i] == false;
bool failpls1 = r.Pls1Result != null && okpls1[i] == false;
bool failann = r.AnnResult != null && okann[i] == false;
//修改修改权重
if (failid && (r.FitResult != null || r.Pls1Result != null || r.AnnResult != null))
{
r.IdWeight = double.NaN;
}
if (failfit && (okid[i] || okpls1[i] || okann[i] || r.Pls1Result != null || r.AnnResult != null))
{
r.FitWeight = double.NaN;
}
if (failpls1 && (okid[i] || okfit[i] || okann[i]))
{
r.Pls1Weight = double.NaN;
}
if (failann && (okid[i] || okfit[i] || okpls1[i]))
{
r.ANNWeight = double.NaN;
}
//求总合
double sumweight = double.Epsilon;
if (!double.IsNaN(r.IdWeight))
{
sumweight += r.IdWeight;
}
if (!double.IsNaN(r.FitWeight))
{
sumweight += r.FitWeight;
}
if (!double.IsNaN(r.Pls1Weight))
{
sumweight += r.Pls1Weight;
}
if (!double.IsNaN(r.ANNWeight))
{
sumweight += r.ANNWeight;
}
//乘系统
if (!double.IsNaN(r.IdWeight))
{
r.IdWeight = r.IdWeight * (100 / sumweight);
}
if (!double.IsNaN(r.FitWeight))
{
r.FitWeight = r.FitWeight * (100 / sumweight);
}
if (!double.IsNaN(r.Pls1Weight))
{
r.Pls1Weight = r.Pls1Weight * (100 / sumweight);
}
if (!double.IsNaN(r.ANNWeight))
{
r.ANNWeight = r.ANNWeight * (100 / sumweight);
}
if (r.IdWeight < 0.00001)
{
r.IdWeight = double.NaN;
}
if (r.FitWeight < 0.00001)
{
r.FitWeight = double.NaN;
}
if (r.Pls1Weight < 0.00001)
{
r.Pls1Weight = double.NaN;
}
if (r.ANNWeight < 0.00001)
{
r.ANNWeight = double.NaN;
}
double sum = 0;
if (!double.IsNaN(r.IdWeight))
{
sum += r.IdValue * r.IdWeight;
}
if (!double.IsNaN(r.FitWeight))
{
sum += r.FitValue * r.FitWeight;
}
if (!double.IsNaN(r.Pls1Weight))
{
sum += r.Pls1Value * r.Pls1Weight;
}
if (!double.IsNaN(r.ANNWeight))
{
sum += r.ANNValue * r.ANNWeight;
}
r.Comp.PredictedValue = sum / 100;//转换百分比
result.Add(r);
}
return(result);
}
public IntegratePropertyResult Predict(Spectrum spec, bool needFilter = true)
{
IntegratePropertyResult result = new IntegratePropertyResult()
{
FitRate = this.FitRate,
IDRate = this.IDRate,
PLS1Rate = this.PLS1Rate,
PLSANNRate = this.PLSANNRate,
Comp = this._comp.Clone()
};
if (this._comp == null)//性质为空，则返回NULL
{
return(null);
}
if (this._pls1 != null && this._pls1.Comp.Name == this._comp.Name)
{
result.PLS1Result = this._pls1.Predict(spec, needFilter);
}
if (this._plsann != null && this._plsann.Comp.Name == this._comp.Name)
{
result.PLSANNResult = this._plsann.Predict(spec, needFilter);
}
//识别
if (this._identify != null && this._identify.Count > 0)
{
foreach (var i in this._identify)
{
if (i.SpecLib.Components.Contains(this._comp.Name))
{
result.IDResult = BindModel.CombineIdResult(result.IDResult, i.Predict(spec, needFilter));
}
}
//过滤其它性质
ComponentList clst = new ComponentList();
foreach (var c in result.IDResult.Components)
{
if (c.Name == this._comp.Name)
{
clst.Add(c.Clone());
break;
}
}
foreach (var s in result.IDResult.Items)
{
var c = s.Spec.Components[this._comp.Name].Clone();
s.Spec.Components = new ComponentList();
s.Spec.Components.Add(c);
}
result.IDResult.Components = clst;
}
//拟合
if (this._fittings != null && this._fittings.Count > 0)
{
var fitmodel = Serialize.DeepClone <FittingModel>(this._fittings.First());
for (int i = 1; i < this._fittings.Count; i++)
{
if (this._fittings[i].SpecLib.Components.Contains(this._comp.Name))
{
fitmodel.SpecLib.Merger(this._fittings[i].SpecLib);
}
}
result.FitResult = fitmodel.Predict(spec, needFilter);
//过滤其它性质
ComponentList clst = new ComponentList();
foreach (var c in result.FitResult.FitSpec.Components)
{
if (c.Name == this._comp.Name)
{
clst.Add(c.Clone());
break;
}
}
result.FitResult.FitSpec.Components = clst;
foreach (var s in result.FitResult.Specs)
{
var c = s.Spec.Components[this._comp.Name].Clone();
s.Spec.Components = new ComponentList();
s.Spec.Components.Add(c);
}
}
return(result);
}
