MLPlatform/date_feature/parameter.yaml

feature_engineering_methods:

  LabelEncoder:
    description: "将分类标签编码为整数。"
    parameters: []

  KBinsDiscretizer:
    description: "将连续数据分箱为离散数据。"
    parameters:
      - name: "n_bins"
        type: "int or array-like"
        default: 5
        description: "指定每个特征要分成的箱数。可以是单个整数，表示所有特征使用相同的箱数；也可以是形状为 (n_features,) 的数组，为每个特征指定不同的箱数。"
      - name: "encode"
        type: "str"
        default: "onehot"
        description: "指定离散化后输出的编码方式。可选值包括 'onehot'（独热编码）、'onehot-dense'（密集独热编码）、'ordinal'（序数编码）。"
      - name: "strategy"
        type: "str"
        default: "quantile"
        description: "定义分箱策略。可选值包括 'uniform'（均匀分箱）、'quantile'（分位数分箱）、'kmeans'（K-Means 聚类分箱）。"

  FunctionTransformer:
    description: "对数据应用自定义函数进行转换。"
    parameters:
      - name: "func"
        type: "callable"
        default: null
        description: "要应用于输入数据的函数。"
      - name: "inverse_func"
        type: "callable"
        default: null
        description: "func 的逆函数，如果存在。"
      - name: "validate"
        type: "bool"
        default: false
        description: "指示是否在转换前验证输入数据。"
      - name: "accept_sparse"
        type: "bool"
        default: false
        description: "指示是否接受稀疏矩阵作为输入。"
      - name: "check_inverse"
        type: "bool"
        default: true
        description: "指示在适合期间是否检查 func 和 inverse_func 是否互为逆函数。"
      - name: "kw_args"
        type: "dict"
        default: null
        description: "传递给 func 的其他关键字参数。"
      - name: "inv_kw_args"
        type: "dict"
        default: null
        description: "传递给 inverse_func 的其他关键字参数。"

  PowerTransformer:
    description: "对数据进行幂变换以使其更符合正态分布。"
    parameters:
      - name: "method"
        type: "str"
        default: "yeo-johnson"
        description: "指定变换方法。可选值包括 'yeo-johnson' 和 'box-cox'。"
      - name: "standardize"
        type: "bool"
        default: true
        description: "指示是否在变换后将数据标准化为零均值和单位方差。"
      - name: "copy"
        type: "bool"
        default: true
        description: "指示是否复制输入数据，或在原地进行变换。"

  QuantileTransformer:
    description: "将数据转换为均匀分布或正态分布。"
    parameters:
      - name: "n_quantiles"
        type: "int"
        default: 1000
        description: "用于分位数变换的分位数数量。"
      - name: "output_distribution"
        type: "str"
        default: "uniform"
        description: "指定输出分布。可选值包括 'uniform' 和 'normal'。"
      - name: "ignore_implicit_zeros"
        type: "bool"
        default: false
        description: "指示是否忽略隐式零。"
      - name: "subsample"
        type: "int"
        default: 100000
        description: "用于计算分位数的子样本大小。"
      - name: "random_state"
        type: "int or None"
        default: null
        description: "用于随机数生成的种子。"
      - name: "copy"
        type: "bool"
        default: true
        description: "指示是否复制输入数据，或在原地进行变换。"

  FeatureHasher:
    description: "使用哈希技巧将特征映射到向量。"
    parameters:
      - name: "n_features"
        type: "int"
        default: 1048576
        description: "哈希空间的维度。"
      - name: "input_type"
        type: "str"
        default: "dict"
        description: "输入数据的类型。可选值包括 'dict' 和 'pair'。"
      - name: "dtype"
        type: "type"
        default: "float64"
        description: "输出数据的类型。"
      - name: "alternate_sign"
        type: "bool"
        default: true
        description: "指示是否在哈希时使用交替符号。"

  DictVectorizer:
    description: "将符号表示的特征（如字典）转换为稀疏矩阵。"
    parameters:
      - name: "dtype"
        type: "type"
        default: "float64"
        description: "输出数据的类型。"
      - name: "separator"
        type: "str"
        default: "="
        description: "用于分隔特征名称的分隔符。"
      - name: "sparse"
        type: "bool"
        default: true
        description: "指示是否返回稀疏矩阵。"
      - name: "sort"
        type: "bool"
        default: true
        description: "指示是否对特征名称排序。"


  PCA:
    description: "主成分分析，用于降维。"
    parameters:
      - name: "n_components"
        type: "int, float, None or str"
        default: null
        description: "要保留的主成分数量。可以是整数、浮点数或 'mle'。"
      - name: "copy"
        type: "bool"
        default: true
        description: "指示是否复制输入数据，或在原地进行变换。"
      - name: "whiten"
        type: "bool"
        default: false
        description: "指示是否对主成分进行白化。"
      - name: "svd_solver"
        type: "str"
        default: "auto"
        description: "用于计算 SVD 的方法。可选值包括 'auto'、'full'、'arpack' 和 'randomized'。"
      - name: "tol"
        type: "float"
        default: 0.0
        description: "奇异值分解的容差。"
      - name: "iterated_power"
        type: "int or 'auto'"
        default: 'auto'
        description: "用于随机化 SVD 的迭代次数。"
      - name: "random_state"
        type: "int or None"
        default: null
        description: "用于随机数生成的种子。"

  SelectKBest:
    description: "选择最重要的 K 个特征。"
    parameters:
      - name: "score_func"
        type: "callable"
        default: "f_classif"
        description: "用于计算特征得分的函数。"
      - name: "k"
        type: "int"
        default: 10
        description: "要选择的特征数量。"

  RFE:
    description: "递归特征消除，用于选择最重要的特征。"
    parameters:
      - name: "estimator"
        type: "object"
        default: null
        description: "用于特征选择的基模型。"
      - name: "n_features_to_select"
        type: "int"
        default: null
        description: "要选择的特征数量。"
      - name: "step"
        type: "int"
        default: 1
        description: "每次迭代要移除的特征数量。"
      - name: "verbose"
        type: "int"
        default: 0
        description: "控制冗长模式的整数。"

  PolynomialFeatures:
    description: "生成多项式特征，增加模型的非线性能力。"
    parameters:
      - name: "degree"
        type: "int"
        default: 2
        description: "生成多项式特征的最高次数。"
      - name: "interaction_only"
        type: "bool"
        default: false
        description: "指示是否仅包含交互项。"
      - name: "include_bias"
        type: "bool"
        default: true
        description: "指示是否包含偏置列。"
      - name: "order"
        type: "str"
        default: "C"
        description: "输出特征的顺序。可选值包括 'C' 和 'F'。"

  OneHotEncoder:
    description: "将分类特征转换为独热编码。"
    parameters:
      - name: "categories"
        type: "str or list or 'auto'"
        default: "auto"
        description: "指定每个特征的类别。"
      - name: "drop"
        type: "str or array-like"
        default: null
        description: "指定要从每个特征中删除的类别。"
      - name: "sparse"
        type: "bool"
        default: true
        description: "指示是否返回稀疏矩阵。"
      - name: "dtype"
        type: "type"
        default: "float64"
        description: "输出数据的类型。"
      - name: "handle_unknown"
        type: "str"
        default: "error"
        description: "指定如何处理未知类别。可选值包括 'error'（抛出异常）、'ignore'（忽略）。"
      - name: "max_categories"
        type: "int or None"
        default: null
        description: "在类别过多时，将类别限制为最大类别数量。"