data structure of a model-OpenSearch(Open Search)-阿里云帮助中心

The Model object represents a machine learning model attached to an OpenSearch application. Models drive ranking and recommendation features—such as popularity scoring, category prediction, and drop-down suggestions—by training on behavioral data collected from your application.

Example

{
    "id": 113023,
    "groupId": "100297752",
    "groupName": "appGroupName",
    "type": "pop",
    "name": "pop_1212",
    "trainTarget": "ctr",
    "cron": "15 0 */2 * *",
    "cronEnabled": true,
    "behaviorEnabled": true,
    "behaviorFromGroupName": "DemoModelName",
    "lastTrainingTime": 1543233439000,
    "auc": 0.85,
    "status": "train_success",
    "progress": 69,
    "isAlreadyDeployed": true,
    "industry": "general",
    "filter": "user_id=1,level=1",
    "fields": [
        {
            "name": "item_title",
            "appFieldType": "title"
        },
        {
            "name": "item_content"
        },
        {
            "name": "item_title_keep",
            "processType": "reserve"
        }
    ],
    "extend": {
        "useHotQuery": true,
        "useHistoryQuery": true
    }
}

Structure

Configurable fields

These fields can be set when creating or updating a model.

Field	Type	Description
`name`	String	The name of the model. Must match the regular expression `/^(?!ops_)[a-zA-Z]\w{0,29}$/`.
`type`	String	The model type. Valid values: `pop` (popularity model), `cp` (category prediction model), `hot` (top search model), `hint` (hint model), `suggest` (drop-down suggestions model).
`industry`	String	The industry the model applies to. Valid values: `general`, `ecommerce`, `content`.
`trainTarget`	String	The training goal. Valid values: `click`, `buy`, `cart`, `collect`, `like`, `comment`, `share`, `subscribe`, `gift`, `download`, `read`, `tip`.
`behaviorEnabled`	Boolean	Specifies whether to use behavioral data for training. Default value: `true`. Not all model types support behavioral data.
`behaviorFromGroupName`	String	The name of the application whose behavioral data is used. Defaults to the current application.
`cron`	String	The cron expression that schedules training runs. For cron syntax, see the Linux Crontab documentation.
`cronEnabled`	Boolean	Specifies whether scheduled training is enabled. Default value: `true`.
`availableThreshold`	Float	The threshold when the model is qualified. Valid values: `0` to `1`.
`filter`	String	A filter expression applied during data import. Supported operators: `<`, `>`, `=`, `!=`, `>=`, `<=`. Separate multiple conditions with commas.
`fields[]`	Object	The application fields the model uses for training.
`fields[].name`	String	The name of the application field.
`fields[].appFieldType`	String	The role of the field in the application. Valid values: `pk` (primary key), `cate_id` (category ID), `title` (title), `cate_name` (category name), `normal` (regular field).
`fields[].processType`	String	The processing type for the field. Valid values: `reserve`, `normal`.
`extend`	Object	Extended configuration for the model.
`extend.useHotQuery`	Boolean	Specifies whether to use hot query-supported models. We recommend that you use the models.
`extend.useHistoryQuery`	Boolean	Specifies whether to use historical query-supported models. We recommend that you use the models.

Read-only fields

These fields are set by the system and cannot be modified directly.

Field	Type	Description
`id`	Integer	The ID of the model.
`groupId`	String	The ID of the application the model belongs to.
`groupName`	String	The name of the application the model belongs to.
`lastTrainingTime`	Integer	The most recent time when the training was complete.
`lastModifyTime`	Integer	The most recent time when the model was modified.
`auc`	Float	The area under curve (AUC) value of the model.
`status`	String	The current status of the model. See Model status values.
`progress`	Integer	The completion percentage of the current training task.
`isAlreadyDeployed`	Boolean	Specifies whether the model has been deployed, or whether one or more models with the same configuration are already available.

Model status values

Status	Description
`train_init`	The model is to be trained.
`train_pending`	The model is being trained.
`validate_failed`	An exception related with the data occurred.
`train_failed`	The training failed.
`train_bad_model`	The training succeeded but the model failed the training.
`train_success`	The training succeeded and the model passed the training.