Get outlier

v1.0.0

Identify outliers (single and repetitive spikes, seasonality, masked outliers, trend and level jumps, amongst other topics) and use for cleansing of the history stream prior to forecast claculation. Depending on math model used, this approach often improves results dramatically, as it removes disturbances.

Input

{
  "type": "object",
  "properties": {
    "Token": {
      "type": "string"
    },
    "data": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "historyValues": {
            "type": "array",
            "items": {
              "type": "number",
              "format": "double"
            }
          },
          "timeSeriesId": {
            "type": "string"
          }
        },
        "additionalProperties": false
      }
    },
    "planningLevelId": {
      "type": "string"
    },
    "startDate": {
      "type": "string"
    }
  },
  "required": [
    "planningLevelId",
    "startDate"
  ],
  "additionalProperties": false
}

Output

{
  "type": "array",
  "items": {
    "type": "object",
    "properties": {
      "outliers": {
        "type": "array",
        "items": {
          "type": "object",
          "properties": {
            "index": {
              "type": "integer",
              "format": "int32"
            },
            "type": {
              "type": "string",
              "enum": [
                "Additive",
                "LevelShift",
                "TemporaryChange",
                "Seasonal"
              ]
            },
            "value": {
              "type": "number",
              "format": "double"
            }
          },
          "additionalProperties": false
        }
      },
      "timeSeriesId": {
        "type": "string"
      }
    },
    "additionalProperties": false
  }
}