Understanding ConfusionMatrix for Google Prediction API

I'm trying to analyze my training model, Google Prediction API provides analyze method to get insights for the model. Currently I want to improve confidence levels, for my predictions, I haven't found any documentation in how to read this ConfusionMatrix, any insights will be great:

{
 "kind": "prediction#analyze",
 "id": "modelX",
 "selfLink": "https://www.googleapis.com/prediction/v1.6/projects/projectX/trainedmodels/modelX/analyze",
 "dataDescription": {
  "outputFeature": {
   "text": [
    {
     "value": "labelA",
     "count": "681"
    },
    {
     "value": "labelB",
     "count": "127"
    },
    {
     "value": "labelC",
     "count": "814"
    },
    {
     "value": "labelD",
     "count": "427"
    }
   ]
  },
  "features": [
   {
    "index": "0",
    "text": {
     "count": "2049"
    }
   }
  ]
 },
 "modelDescription": {
  "modelinfo": {
   "kind": "prediction#training"
  },
  "confusionMatrix": {
   "labelA": {
    "labelA": "14.17",
    "labelB": "0.17",
    "labelC": "3.83",
    "labelD": "0.67"
   },
   "labelB": {
    "labelA": "0.50",
    "labelB": "2.00",
    "labelC": "1.33",
    "labelD": "0.00"
   },
   "labelC": {
    "labelA": "1.17",
    "labelB": "0.00",
    "labelC": "70.00",
    "labelD": "3.50"
   },
   "labelD": {
    "labelA": "1.17",
    "labelB": "0.17",
    "labelC": "4.17",
    "labelD": "12.17"
   }
  },
  "confusionMatrixRowTotals": {
   "labelA": "18.83",
   "labelB": "3.83",
   "labelC": "74.67",
   "labelD": "17.67"
  }
 }
}