All Products
Search

This Product

    Platform For AI:Swing Train

    Document Center

    Platform For AI:Swing Train

    Last Updated:Jun 02, 2026

    Swing is an item recall algorithm. Use the Swing Train component to measure item similarity based on user-item-user principles.

    Limits

    Supported compute engines: MaxCompute and Realtime Compute for Apache Flink.

    Configure the component

    Configure the component by using one of the following methods:

    Method 1: Configure the component in Machine Learning Designer

    Configure the following parameters on the pipeline configuration tab.
    Tab Parameter Description
    Field Setting itemCol Name of the item column.
    userCol Name of the user column.
    Parameter Setting alpha Smoothing factor. Default value: 1.0.
    maxItemNumber Maximum number of users per item for calculation. Default value: 1000.
    Note If an item exceeds this count, the algorithm randomly samples this number of users.
    maxUserItems Maximum number of items per user for calculation. Default value: 1000.
    Note Users exceeding this item count are excluded from calculation.
    minUserItems Minimum number of items per user for calculation. Default value: 10.
    Note Users below this item count are excluded from calculation.
    resultNormalize Whether to normalize results.
    userAlpha Alpha parameter for users. Default value: 5.0.
    userBeta Beta parameter for users. Default value: -0.35.
    Execution Tuning Number of Workers Number of worker nodes. Must be used with Memory per worker, unit MB. Valid values: 1 to 9999.
    Memory per worker, unit MB Memory per worker node, in MB. Valid values: 1024 to 65536.

    Method 2: Use Python code

    Configure component parameters with Python code. Run the code with the Python Script V2 component. For more information, see Python Script.
    Parameter Required Description Default value
    itemCol Yes Name of the item column. None
    userCol Yes Name of the user column. None
    alpha No Alpha smoothing factor. 1.0
    userAlpha No Alpha parameter for users.
    Note Used to calculate user weight: weight = 1.0/(userAlpha + userClickCount)^userBeta.
    		5.0
    userBeta No Beta parameter for users.
    Note Used to calculate user weight: weight = 1.0/(userAlpha + userClickCount)^userBeta.
    		-0.35
    resultNormalize No Whether to normalize results. false
    maxItemNumber No Maximum number of users per item for calculation.
    Note If an item exceeds this count, the algorithm randomly samples this number of users.
    		1000
    minUserItems No Minimum number of items per user for calculation.
    Note Users below this item count are excluded from calculation.
    		10
    maxUserItems No Maximum number of items per user for calculation.
    Note Users exceeding this item count are excluded from calculation.
    		1000
    Sample code: 
    df_data = pd.DataFrame([
    ["a1", "11L", 2.2],
    ["a1", "12L", 2.0],
    ["a2", "11L", 2.0],
    ["a2", "12L", 2.0],
    ["a3", "12L", 2.0],
    ["a3", "13L", 2.0],
    ["a4", "13L", 2.0],
    ["a4", "14L", 2.0],
    ["a5", "14L", 2.0],
    ["a5", "15L", 2.0],
    ["a6", "15L", 2.0],
    ["a6", "16L", 2.0],
])

data = BatchOperator.fromDataframe(df_data, schemaStr='user string, item string, rating double')


model = SwingTrainBatchOp()\
    .setUserCol("user")\
    .setItemCol("item")\
    .setMinUserItems(1)\
    .linkFrom(data)

model.print()

predictor = SwingRecommBatchOp()\
    .setItemCol("item")\
    .setRecommCol("prediction_result")

predictor.linkFrom(model, data).print()

    Example

    The following figure shows a sample Swing Train pipeline. ExampleConfigure the components as follows:
    1. Prepare a training dataset and a test dataset.
    2. Create two MaxCompute tables: Table 1 with userid and itemid fields, and Table 2 with the itemid field. All fields are STRING type. Upload the training dataset to Table 1 and the test dataset to Table 2 by using the tunnel command. Set Read Table-1 to Table 1 and Read Table-2 to Table 2. For more information about how to install and configure the MaxCompute client, see Connect using the local client (odpscmd). For more information about Tunnel commands, see Tunnel commands.
    3. Import the training dataset to the Swing Train component and configure the parameters as described in Method 1: Configure the component in Machine Learning Designer.
    4. Import the test dataset and the model to the Swing prediction component for prediction.