最大模式长度 fpGrowth (Apache) PySpark

Question

我正在尝试运行使用 PySpark 的关联规则。我首先创建一个 FPGrowth 树并将其传递给关联规则方法。

但是，我希望添加一个最大模式长度参数，以限制我想要在左轴和右轴上显示的项目数。对于项目之间的关联，我只想将模式长度保持为 2。

## fit model

from pyspark.ml.fpm import FPGrowth

fpGrowth_1 = FPGrowth(itemsCol="collect_set(title_name)", minSupport=.001, minConfidence=0.001)

model_working_1 = fpGrowth_1.fit(transactions_2)

## Display frequent itemsets.

model_working_1.freqItemsets.show()

+--------------------+------+
|               items|  freq|
+--------------------+------+
|[Temptation Islan...|325291|
|[Temptation Island] |282205|
|[Temptation Islan...|175694|
|[S4 - Engl  progr...|171400|
|      [Nieuwe Buren]|168684|
|[Neighboursss, Te...|113113|
|       [Love Island]|146766|
|[Love Island, S4 ...| 65285|
|[Love Island, Tem...|105834|
|[Love Island, Tem...| 83335|
|[Love Island, Tem...|115979|
|[Good Time Sle......|132439|
+--------------------+------+

# Display generated association rules.
model_working_1.associationRules.show()
+--------------------+--------------------+------------------+
|          antecedent|          consequent|        confidence|
+--------------------+--------------------+------------------+
|[Love Island, Tem...| [Temptation Island]|0.7185352520714957|
|[De Beste Verleid...|[Temptation Islan...|0.9147820487266372|
|     [Bella Donna's]|[Temptation Islan...|  0.74988107580655|
|[Binnenkort bij V...|[Temptation Islan...|0.9756179956817415|
|[Married at First...| [Temptation Island]|0.8692627446452283|
|       [Love Island]| [Temptation Island]|0.7211070683945873|
|       [Love Island]|[Temptation Islan...|0.7902307073845442|
|[S4 - Dutch progr...| [Temptation Island]|  0.61975495915986|
|[S4 - Dutch progr...|[Temptation Islan...|0.7550758459743291|
|[The Good Doctor,...| [Temptation Island]| 0.873575189492565|
+--------------------+--------------------+------------------+


# transform examines the input items against all the association rules and summarize the

# consequents as prediction

model_working_1.transform(transactions_2).show()

+---------------------+----------------------------------------------------------------------------------------------+
|         title_name  |        Prediction                                                        |        
+---------------------+----------------------------------------------------------------------------------------------+
|[Goode Time Bad  ....| Temptation Island VIPS,S4 - Dutch program viewer,Weg van Jou                                                                                          |  
                                                     The Good Doctor,Moordvrouw,De 12 van Oldenheim,Married at First Sight,Dave en Dien op Ibiza,Temptation Gossip]                           |  
|[S4 - Englis progr...|Lara Croft Tomb Raider, Ronald Goedemondt - Geen Sp
|[Goede Tijden Sl.........|[I Love You Tattoo, S7 - Dutch suspense-series viewer, Temptation Island VIPS, Awkward, Goede Tijden Slechte Tijden, Lost, De Beste Verleiders, Cellblock H]|

生成的关联规则是非常长的模式。我想将长度保持在 2 个模式或更多一点。现在我要解释或理解太多了。

有什么方法可以限制 PySPark 中的模式长度？我为 scala 找到了一个 link 但在 PySaprk 中没有这样的东西。

如果你能在这种情况下 suggest/help 我将不胜感激。提前致谢！！！

Answer 1

在pyspark你可以试试：

from pyspark.sql.functions import col, size
model.associationRules.where(size(col('antecedent')) == 1).where(size(col('cosequent')) == 1).show()

最大模式长度 fpGrowth (Apache) PySpark

Maximum Pattern Length fpGrowth (Apache) PySpark

apache-spark

pyspark

fpgrowth