Skip to content

4.validate.md

Latest commit

 

History

History
305 lines (280 loc) · 7.92 KB

4.validate.md

File metadata and controls

305 lines (280 loc) · 7.92 KB
Raw

4. mljs validate

  1. According to the probability and sort and unique the data. Caculate the between the value threshold.
  2. determine the label by the thresthold.
  3. according to label in fact and the label predict.
var config_model='';
var data=[];
for(var i=0;i<trainingData.length;i++)
  {
    var row=[];
    row.push(trainingData[i][0]);
    var name='model'+i;
    window[name]=train_model_new(trainingData[i][1]);
    //model=
    config_model+='<p>'+trainingData[i][0]+" "+window[name].weight+" intercept"+window[name].intercept+'</p>';
    proarr=model_data_probability(window[name],validateData[i][1]);
    rocarray=mljs_validate(proarr,validateLabel[i][1]);
    row.push(rocarray[3]);
    row.push(rocarray[1]);
    row.push(rocarray[2]);
    //console.log(row);
    data.push(row);
  }
  //console.log(data);
  plot_roc(data,'roc_plot');

mljs_validate

function mljs_validate(probability,trainingLabel){
  var cutoffarray=cutoff(probability);
//console.log(probability);
//console.log(cutoffarray);
roc_array_label=mljs_determin_label(probability,cutoffarray);
//console.log(roc_array_label);
roc_tp4=mljs_label_cross(trainingLabel,roc_array_label);
//console.log(roc_array_label);
return roc_tp4;
//console.log(trainingLabel);
// var data=[['gene',roc_tp4[3],roc_tp4[1],roc_tp4[2]]];
//    plot_roc(data,'roc_plot');

}

1. According to the probability and sort and unique the data. Caculate the between the value threshold.
cutoff(probability);

  1. Deep copy probability array.
  2. sort the probability from small to big.
  3. Unique the probability.
  4. Calculate the therthold the first one is the smallest -1, other is the (data[i]+data[i+1])/2
var cutoffarray=cutoff(probability);
function cutoff(arr2) {
  arr=JSON.parse(JSON.stringify(arr2));
  if (arr.length === 0) return arr;
  arr = arr.sort(function (a, b) { return a*1 - b*1; });
  var ret = [arr[0]];
  for (var i = 1; i < arr.length; i++) { //Start loop at 1: arr[0] can never be a duplicate
    if (arr[i-1] !== arr[i]) {
      ret.push(arr[i]);
    }
  }
  var cutoff=[];
  cutoff.push(ret[0]-1);
  for(var i=1;i<ret.length;i++){
    cutoff.push((ret[i-1]+ret[i])/2);
  }
  cutoff.push(ret[ret.length-1]+1);
  return cutoff;
}

2. determine the label by the thresthold.
mljs_determin_label(probability,cutoffarray)

roc_array_label=mljs_determin_label(probability,cutoffarray);
function mljs_determin_label(a,b){
  //a is probability array
  //b is cutoff array
  label_array=[];
  for(var i=0;i<b.length;i++){
    //get each cutoff valule is b[i]
    label_array.push([]);
    for(var j=0;j<a.length;j++){
      //get the each probability a[j]
      if(a[j]> b[i]){
        label_array[i].push(1);
      }else{
        label_array[i].push(0);
      }
    }
  }
return label_array;
}

3. according to label in fact and the label predict.
mljs_label_cross(trainingLabel,roc_array_label);

roc_tp4=mljs_label_cross(trainingLabel,roc_array_label);
function mljs_label_cross(a,b){
  //a is the label of fact.
  //b is the array of use probability to predict label.
  var pro=[];
  var tprarr=[];//sensitive
  var fprarr=[];//
  var precisionarr=[];
  var specificityarr=[];
  for(var i=0;i<b.length;i++){
    // each cut off label is b[i]
    var tp=0;
    var tf=0;
    var fp=0;
    var fn=0;
    var tpr=0;
    var fpr=0;
    var precision=0;
    var specificity=0;
    for(var j=0;j<b[i].length;j++){
      if(b[i][j]==1 & a[j]==1){
        tp++;
      }
      if(b[i][j]==0 & a[j]==0){
        tf++;
      }
      if(b[i][j]==0 & a[j]==1){
        fn++;
      }
      if(b[i][j]==1 & a[j]==0){
        fp++;
      }
    }
    tpr=tp/(tp+fn);//sensitive
    tpr.toFixed(3);
    fpr=fp/(tf+fp);
    fpr.toFixed(3);
    precision=tp/(tp+fp);
    precision.toFixed(3);
    specificity=tf/(tf+fp);
    specificity.toFixed(3);

    pro.push([tp,tf,fp,fn]);
    fprarr.push(fpr);
    tprarr.push(tpr);
    precisionarr.push(precision);
    specificityarr.push(specificity);
  }
  var auc=0;
  for(var i=1;i<fprarr.length;i++){
    auc+=(tprarr[i]+tprarr[i-1])*(fprarr[i]-fprarr[i-1])/2;
  }
  auc=Math.abs(auc).toFixed(3);
  return [pro,fprarr,tprarr,auc,precisionarr,specificityarr];
}
//output is [[tp,tf,fp,fn],[false positive ratio array],[sensitive array],auc,[precision array],[specificity array]]

//data=[[tp,tf,fp,fn],[false positive ratio array],[sensitive array],auc,[precision array],[specificity array]]
function report_ml_value(data)
{
var sensitive=data[2];
var auc=data[3];
var precision=data[4];
var specificity=data[5];
//first get the sensitive not less than 0.9.
var sen_09=0;
for(var i=0;i<sensitive.length;i++)
{
  if(sensitive[i]<0.9)
  {
    sen_09=i-1;
    break;
  }
}
  console.log(sensitive[sen_09]+' '+precision[sen_09]+' '+specificity[sen_09]);
var spe_09=0;
for(var i=0;i<specificity.length;i++)
{
  if(specificity[i]>=0.9)
  {
    spe_09=i;
    break;
  }
}
  console.log(sensitive[spe_09]+' '+precision[spe_09]+' '+specificity[spe_09]);
return [[sensitive[sen_09],precision[sen_09],specificity[sen_09],auc],[sensitive[spe_09],precision[spe_09],specificity[spe_09],auc]];
}

report_array=report_ml_value(roc_tp4)

export table.

//html_report_value='';
function generate_table_report(name,data)
{
var html_report_value='<table class="table"  id="logistic_report'+name+'"> \
  <thead> \
    <tr>  <th scope="col">Name</th> <th scope="col">Sensitive</th> <th scope="col">Precision</th><th scope="col">Specificity</th><th scope="col">AUC</th>\
    </tr>\
    </thead><body>';
  
  for(var i=0;i<data.length;i++)
  {
    html_report_value+='<tr><td>'+name+'</td><td>'+data[i][0]+'</td><td>'+data[i][1]+'</td><td>'+data[i][2]+'</td><td>'+data[i][3]+'</td></tr>';
  }
  html_report_value+='</body></table>';
  return html_report_value;
}
name='model1';
report_array=report_ml_value(roc_tp4);
generate_table_report(name,report_array);

old version

select_data_from_id()

function select_data_from_id()
{
//get data;
var data= getinforvalue('training_decision');
var label=getinforlabel('label_decision');
var label_unique=sort_unique(label);
//console.log(label_unique);
//merge_matrix()


merge_array=merge_matrix(data,label);
//console.log(merge_array);
merge_array=shufflearray(merge_array);
//console.log(merge_array);
//
var train_range_ratio=[0,1];
var validate_range_ratio=[0,1];
var train_range=[];
var validate_range=[];
var number=parseInt(train_range_ratio[1]*(data.length)) ;

//train_range_ratio[1]*(data.length);
train_range.push(parseInt(train_range_ratio[0]*(data.length)));
train_range.push(parseInt(train_range_ratio[1]*(data.length)));
//
validate_range.push(parseInt(validate_range_ratio[0]*(data.length)));
validate_range.push(parseInt(validate_range_ratio[1]*(data.length)));

//
var train=generate_train_validate_data(label_unique,train_range,merge_array);
trainingData =train[0];
trainingLabel=train[1];
var valide=generate_train_validate_data(label_unique,validate_range,merge_array);
validateData=valide[0];
validateLabel=valide[1];

/*
model=train_model(trainingData[0][1]);
proarr=model_data_probability(model,validateData[0][1]);
rocarray=mljs_validate(proarr,validateLabel[0][1]);
console.log(rocarray);
*/
var data=[];
for(var i=0;i<trainingData.length;i++)
  {
    var row=[];
    row.push(trainingData[i][0]);
    model=train_model(trainingData[i][1]);
    proarr=model_data_probability(model,validateData[i][1]);
    rocarray=mljs_validate(proarr,validateLabel[i][1]);
    row.push(rocarray[3]);
    row.push(rocarray[1]);
    row.push(rocarray[2]);
    //console.log(row);
    data.push(row);
  }
  console.log(data);
  plot_roc(data,'roc_plot');

}
rocarray=mljs_validate(proarr,validateLabel[0][1]);
function mljs_validate(probability,trainingLabel){
  var cutoffarray=cutoff(probability);
console.log(probability);
console.log(cutoffarray);
roc_array_label=mljs_determin_label(probability,cutoffarray);
roc_tp4=mljs_label_cross(trainingLabel,roc_array_label);
return roc_tp4;
}