使用量
microsoftml.rx_featurize(data: typing.Union[revoscalepy.datasource.RxDataSource.RxDataSource,
pandas.core.frame.DataFrame],
output_data: typing.Union[revoscalepy.datasource.RxDataSource.RxDataSource,
str] = None, overwrite: bool = False,
data_threads: int = None, random_seed: int = None,
max_slots: int = 5000, ml_transforms: list = None,
ml_transform_vars: list = None, row_selection: str = None,
transforms: dict = None, transform_objects: dict = None,
transform_function: str = None,
transform_variables: list = None,
transform_packages: list = None,
transform_environment: dict = None, blocks_per_read: int = None,
report_progress: int = None, verbose: int = 1,
compute_context: revoscalepy.computecontext.RxComputeContext.RxComputeContext = None)
說明
將輸入資料集轉換成輸出資料集。
論點
資料
例如 revoscalepy 的資料來源物件、資料框架,或檔案 .xdf 的路徑。
output_data
輸出文字或 xdf 檔名,或 RxDataSource 具備寫入功能以儲存轉換後的資料。 若 為 N,則回傳一個資料框。 預設值是 None。
覆寫
若 True,output_data則存在被覆寫;若Falseoutput_data存在未被覆寫。 預設值為 False。
data_threads
一個整數,指定資料管線中所需的平行程度。 若 無,則執行緒數量由內部決定。 預設值是 None。
random_seed
指定隨機種子。 預設值是 None。
max_slots
最大回傳向量值欄位的槽位數(<=0 以返回全部)。
ml_transforms
指定一份在訓練前要對資料執行的 MicrosoftML 轉換清單,或是「若不執行轉換則 無 」。 參見 featurize_text、 categorical、 categorical_hash和 ,以了解支援的轉換。
這些轉換會在指定的 Python 轉換後執行。
預設值是 None。
ml_transform_vars
指定一個變數名稱的字元向量, ml_transforms 若不使用變數名稱,則指定為 None 。
預設值是 None。
row_selection
不支援。 指定資料集中的列(觀察值),模型將使用資料集中的邏輯變數名稱(引號內)或使用資料集中變數的邏輯表達式。 例如:
row_selection = "old"僅使用變數old值為True的觀測值。row_selection = (age > 20) & (age < 65) & (log(income) > 10)僅使用變數值age介於20至65之間且變數值logincome大於10的觀測值。
列選擇是在處理任何資料轉換後執行的(參見參數 transforms 或 transform_function)。 與所有表達式一樣, row_selection 可以在函 expression 式呼叫之外使用函式定義。
變換
不支援。 代表第一輪變數轉換的表達式。 與所有表達式一樣, transforms (或 row_selection)可以在函式呼叫之外使用函 expression 式定義。
預設值是 None。
transform_objects
不支援。 一個命名的清單,包含可被 transforms、 transform_function、 row_selection和 所引用的物件。 預設值是 None。
transform_function
變數轉換函數。 預設值是 None。
transform_variables
一個用於轉換函數所需的輸入資料集變數的字元向量。 預設值是 None。
transform_packages
不支援。 一個字元向量,指定額外 Python 套件(除 在 中 RxOptions.get_option("transform_packages")指定的套件外)可供並預載,用於變數轉換函數。
例如,在 revoscalepy 函式中透過其 transforms 和 transform_function 參數明確定義的函數,或是透過 their formula 或 row_selection 參數隱含定義的函式。
transform_packages參數也可能是 None,表示外部沒有RxOptions.get_option("transform_packages")套件被預先載入。
transform_environment
不支援。 一個由使用者定義的環境,作為所有內部開發環境的父環境,並用於變量資料轉換。
如果 transform_environment = None,則會使用新的「雜湊」環境,父 revoscalepy.baseenv 為母節點,預設值為 None。
blocks_per_read
指定從資料來源讀取的每一塊資料要讀取的區塊數量。
report_progress
一個整數值,用來指定對資料列處理進度的報告層級:
0:沒有進展報告。1:已處理的列數會被列印並更新。2:處理的行數及時間報告。3:列數已處理,所有時間點都會被報告。
預設值為 1。
詳細資訊
一個整數值,用來指定想要輸出的量。
若 0,則計算過程中不會印出冗長輸出。 整數值 來自 1 , 4 提供越來越多的資訊。
預設值為 1。
compute_context
設定計算執行的上下文,並以有效的 revoscalepy 指定。RxComputeContext。 目前是局部且 revoscalepy。支援 RxInSqlServer 的運算上下文。
Returns
資料框架或 revoscalepy。RxDataSource 物件代表所建立的輸出資料。
範例
'''
Example with rx_featurize.
'''
import numpy
import pandas
from microsoftml import rx_featurize, categorical
# rx_featurize basically allows you to access data from the MicrosoftML transforms
# In this example we'll look at getting the output of the categorical transform
# Create the data
categorical_data = pandas.DataFrame(data=dict(places_visited=[
"London", "Brunei", "London", "Paris", "Seria"]),
dtype="category")
print(categorical_data)
# Invoke the categorical transform
categorized = rx_featurize(data=categorical_data,
ml_transforms=[categorical(cols=dict(xdatacat="places_visited"))])
# Now let's look at the data
print(categorized)
輸出:
places_visited
0 London
1 Brunei
2 London
3 Paris
4 Seria
Beginning processing data.
Rows Read: 5, Read Time: 0, Transform Time: 0
Beginning processing data.
Beginning processing data.
Rows Read: 5, Read Time: 0, Transform Time: 0
Beginning processing data.
Elapsed time: 00:00:00.0521300
Finished writing 5 rows.
Writing completed.
places_visited xdatacat.London xdatacat.Brunei xdatacat.Paris \
0 London 1.0 0.0 0.0
1 Brunei 0.0 1.0 0.0
2 London 1.0 0.0 0.0
3 Paris 0.0 0.0 1.0
4 Seria 0.0 0.0 0.0
xdatacat.Seria
0 0.0
1 0.0
2 0.0
3 0.0
4 1.0