spark ml实现逻辑回归案例分析

一、spark ml介绍

spark ml对机器学习算法的api进行了标准化，使将多个算法合并到一个管道或工作流变得更容易。为了更清楚了解，从以下及几个方面展开说明。

DataFrame:这个ML API使用Spark SQL的DataFrame作为ML数据集，它可以容纳各种数据类型。例如，DataFrame可能有不同的列存储文本、特征向量、真实标签和预测。

Transformer: Transformer是一种可以将一个DataFrame转换成另一个DataFrame的算法。例如，ML模型是一个转换器，它将具有特性的DataFrame转换为具有预测的DataFrame。

Estimator:估计器是一种算法，用于DataFrame转换。例如，学习算法是一种估计器，它训练一个DataFrame并生成一个模型。

pipeline:管道将多个变压器和估计器链接在一起，以指定一个ML工作流。

二、spark ml实现

尝试用spark ml实现广告点击预测，训练和测试数据使用Kaggle Avazu CTR 比赛的样例数据，下载地址：https://www.kaggle.com/c/avazu-ctr-prediction/data。

开发环境：java1.8.0_172+scala2.11.8+spark2.3.1

依赖包

<dependency>

<groupId>org.apache.spark</groupId>

<artifactId>spark-core_2.11</artifactId>

<version>2.3.1</version>

</dependency>

<!-- https://mvnrepository.com/artifact/org.apache.spark/spark-sql -->

<dependency>

<groupId>org.apache.spark</groupId>

<artifactId>spark-sql_2.11</artifactId>

<version>2.3.1</version>

</dependency>

<!-- https://mvnrepository.com/artifact/org.apache.spark/spark-hive -->

<dependency>

<groupId>org.apache.spark</groupId>

<artifactId>spark-hive_2.11</artifactId>

<version>2.3.1</version>

</dependency>

<!-- https://mvnrepository.com/artifact/org.apache.spark/spark-mllib -->

<dependency>

<groupId>org.apache.spark</groupId>

<artifactId>spark-mllib_2.11</artifactId>

<version>2.3.1</version>

</dependency>

1. 数据集

spark加载csv文件，dataframe基本结构如下：

val data = spark.read.csv("/opt/data/ads_6M.csv").toDF(

"id","click","hour","C1","banner_pos","site_id","site_domain",

"site_category","app_id","app_domain","app_category","device_id","device_ip",

"device_model","device_type","device_conn_type","C14","C15","C16","C17","C18",

"C19","C20","C21")

data.show(5,false)

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+---+------+---+

|id |click|hour |C1 |banner_pos|site_id |site_domain|site_category|app_id |app_domain|app_category|device_id|device_ip|device_model|device_type|device_conn_type|C14 |C15|C16|C17 |C18|C19|C20 |C21|

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+---+------+---+

|10153523536315735769|0 |14102100|1005|0 |85f751fd|c4e18dd6 |50e219e0 |53de0284|d9b5648e |0f2161f8 |a99f214a |788c3e75 |2ea4f8ba |1 |0 |20508|320|50 |2351|3 |163|-1 |61 |

|10448041871517116234|0 |14102100|1005|0 |1fbe01fe|f3845767 |28905ebd |ecad2386|7801e8d9 |07d7df22 |a99f214a |99cd8fa2 |81b42528 |1 |0 |15707|320|50 |1722|0 |35 |-1 |79 |

|10488488220071431784|0 |14102100|1005|1 |72a56356|45368af7 |3e814130 |ecad2386|7801e8d9 |07d7df22 |a99f214a |e8fc2f9f |900981af |1 |2 |18993|320|50 |2161|0 |35 |-1 |157|

|10625948582770087788|0 |14102100|1005|0 |85f751fd|c4e18dd6 |50e219e0 |5e3f096f|2347f47a |0f2161f8 |a99f214a |9c1b8be7 |24f6b932 |1 |0 |18993|320|50 |2161|0 |35 |100215|157|

|11151072182888929242|0 |14102100|1005|1 |5b4d2eda|16a36ef3 |f028772b |ecad2386|7801e8d9 |07d7df22 |a99f214a |866e0a54 |d787e91b |1 |0 |16208|320|50 |1800|3 |167|-1 |23 |

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+---+------+---+

包含24个字段：

• 1-id: ad identifier
• 2-click: 0/1 for non-click/click
• 3-hour: format is YYMMDDHH, so 14091123 means 23:00 on Sept. 11, 2014 UTC.
• 4-C1 — anonymized categorical variable
• 5-banner_pos
• 6-site_id
• 7-site_domain
• 8-site_category
• 9-app_id
• 10-app_domain
• 11-app_category
• 12-device_id
• 13-device_ip
• 14-device_model
• 15-device_type
• 16-device_conn_type
• 17~24—C14-C21 — anonymized categorical variables

其中5到15列为分类特征，16~24列为数值型特征。将数据集分为训练集和测试集，比例为0.7：0.3。

val splited = data.randomSplit(Array(0.7,0.3),2L)

2. 特征处理

2.1 StringIndexer

对于分类特征可以使用StringIndexer将标签的字符串列编码为标签索引列，将字符串特征转化为数值特征，便于下游管道组件处理。

val catalog_features = Array("click","site_id","site_domain","site_category","app_id","app_domain","app_category","device_id","device_ip","device_model")

var train_index = splited(0)

var test_index = splited(1)

for(catalog_feature <- catalog_features){

val indexer = new StringIndexer()

.setInputCol(catalog_feature)

.setOutputCol(catalog_feature.concat("_index"))

val train_index_model = indexer.fit(train_index)

val train_indexed = train_index_model.transform(train_index)

val test_indexed = indexer.fit(test_index).transform(test_index,train_index_model.extractParamMap())

train_index = train_indexed

test_index = test_indexed

}

println("字符串编码下标标签：")

train_index.show(5,false)

test_index.show(5,false)

字符串编码下标标签：

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+----+------+---+-----------+-------------+-----------------+-------------------+------------+----------------+------------------+---------------+---------------+------------------+

|id |click|hour |C1 |banner_pos|site_id |site_domain|site_category|app_id |app_domain|app_category|device_id|device_ip|device_model|device_type|device_conn_type|C14 |C15|C16|C17 |C18|C19 |C20 |C21|click_index|site_id_index|site_domain_index|site_category_index|app_id_index|app_domain_index|app_category_index|device_id_index|device_ip_index|device_model_index|

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+----+------+---+-----------+-------------+-----------------+-------------------+------------+----------------+------------------+---------------+---------------+------------------+

|10000133892746881176|0 |14102813|1005|0 |85f751fd|c4e18dd6 |50e219e0 |febd1138|82e27996 |0f2161f8 |a99f214a |f5c62586 |b4b19c97 |1 |0 |21611|320|50 |2480|3 |297 |100111|61 |0.0 |0.0 |0.0 |0.0 |4.0 |4.0 |1.0 |0.0 |23751.0 |64.0 |

|10000987464039884177|0 |14102816|1005|0 |5bcf81a2|9d54950b |f028772b |ecad2386|7801e8d9 |07d7df22 |a99f214a |845f69f4 |fa61e8fe |1 |0 |23438|320|50 |2684|2 |1319|-1 |52 |0.0 |11.0 |7.0 |1.0 |0.0 |0.0 |0.0 |0.0 |5237.0 |67.0 |

|10001055656394300907|0 |14102814|1005|0 |85f751fd|c4e18dd6 |50e219e0 |e9739828|df32afa9 |cef3e649 |a99f214a |6454c6ba |ecb851b2 |1 |0 |23441|320|50 |2685|1 |33 |100083|212|0.0 |0.0 |0.0 |0.0 |13.0 |11.0 |2.0 |0.0 |18147.0 |8.0 |

|10001237608243220141|0 |14102701|1005|0 |85f751fd|c4e18dd6 |50e219e0 |febd1138|82e27996 |0f2161f8 |a99f214a |ab986e15 |2ea4f8ba |1 |0 |19743|320|50 |2264|3 |427 |100000|61 |0.0 |0.0 |0.0 |0.0 |4.0 |4.0 |1.0 |0.0 |23941.0 |34.0 |

|10001363001408225332|0 |14102812|1005|1 |85f751fd|c4e18dd6 |50e219e0 |1dc72b4d|2347f47a |0f2161f8 |b7c2e4b6 |bce45090 |5db079b5 |1 |2 |22998|300|50 |2657|3 |35 |100013|23 |0.0 |0.0 |0.0 |0.0 |25.0 |1.0 |1.0 |1760.0 |729.0 |25.0 |

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+----+------+---+-----------+-------------+-----------------+-------------------+------------+----------------+------------------+---------------+---------------+------------------+

only showing top 5 rows

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+---+------+---+-----------+-------------+-----------------+-------------------+------------+----------------+------------------+---------------+---------------+------------------+

|id |click|hour |C1 |banner_pos|site_id |site_domain|site_category|app_id |app_domain|app_category|device_id|device_ip|device_model|device_type|device_conn_type|C14 |C15|C16|C17 |C18|C19|C20 |C21|click_index|site_id_index|site_domain_index|site_category_index|app_id_index|app_domain_index|app_category_index|device_id_index|device_ip_index|device_model_index|

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+---+------+---+-----------+-------------+-----------------+-------------------+------------+----------------+------------------+---------------+---------------+------------------+

|10002333262420133303|0 |14102211|1005|1 |856e6d3f|58a89a43 |f028772b |ecad2386|7801e8d9 |07d7df22 |a99f214a |ac322dfb |0dc22ebc |1 |0 |19771|320|50 |2227|0 |679|100077|48 |0.0 |6.0 |6.0 |1.0 |0.0 |0.0 |0.0 |0.0 |6004.0 |279.0 |

|10002749335348787004|1 |14102800|1005|0 |2a68aa20|9b851bd8 |3e814130 |ecad2386|7801e8d9 |07d7df22 |a99f214a |b4a0ec64 |49bc419a |1 |0 |20213|320|50 |2316|0 |167|100079|16 |1.0 |57.0 |56.0 |3.0 |0.0 |0.0 |0.0 |0.0 |30.0 |563.0 |

|10003763177308262205|0 |14102814|1002|0 |7971d583|c4e18dd6 |50e219e0 |ecad2386|7801e8d9 |07d7df22 |fffcf8a4 |f615f762 |a5df7413 |0 |0 |23441|320|50 |2685|1 |33 |-1 |212|0.0 |408.0 |0.0 |0.0 |0.0 |0.0 |0.0 |1003.0 |5471.0 |982.0 |

|10005435104591133943|0 |14102719|1005|0 |85f751fd|c4e18dd6 |50e219e0 |92f5800b|ae637522 |0f2161f8 |a99f214a |8f2784a2 |0bcabeaf |1 |3 |21189|320|50 |2424|1 |161|100193|71 |0.0 |0.0 |0.0 |0.0 |1.0 |2.0 |1.0 |0.0 |4207.0 |19.0 |

|10006076676750034840|0 |14102522|1005|1 |e151e245|7e091613 |f028772b |ecad2386|7801e8d9 |07d7df22 |a99f214a |dc88197f |fce66524 |1 |0 |4687 |320|50 |423 |2 |39 |100148|32 |0.0 |2.0 |2.0 |1.0 |0.0 |0.0 |0.0 |0.0 |4109.0 |22.0 |

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+---+------+---+-----------+-------------+-----------------+-------------------+------------+----------------+------------------+---------------+---------------+------------------+

only showing top 5 rows

2.2 FeatureHasher

特征哈希将一组分类或数值特征投射到指定维的特征向量(通常比原始特征空间小很多)。这是使用哈希技巧将特征映射到特征向量中的索引。

val hasher = new FeatureHasher()

.setInputCols("site_id_index","site_domain_index","site_category_index","app_id_index","app_domain_index","app_category_index","device_id_index","device_ip_index","device_model_index","device_type","device_conn_type","C14","C15","C16","C17","C18","C19","C20","C21")

.setOutputCol("feature")

val train_hs = hasher.transform(train_index)

val test_hs = hasher.transform(test_index)

println("特征Hasher编码：")

train_index.show(5,false)

test_index.show(5,false)

特征Hasher编码：

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+----+------+---+-----------+-------------+-----------------+-------------------+------------+----------------+------------------+---------------+---------------+------------------+

|id |click|hour |C1 |banner_pos|site_id |site_domain|site_category|app_id |app_domain|app_category|device_id|device_ip|device_model|device_type|device_conn_type|C14 |C15|C16|C17 |C18|C19 |C20 |C21|click_index|site_id_index|site_domain_index|site_category_index|app_id_index|app_domain_index|app_category_index|device_id_index|device_ip_index|device_model_index|

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+----+------+---+-----------+-------------+-----------------+-------------------+------------+----------------+------------------+---------------+---------------+------------------+

|10000133892746881176|0 |14102813|1005|0 |85f751fd|c4e18dd6 |50e219e0 |febd1138|82e27996 |0f2161f8 |a99f214a |f5c62586 |b4b19c97 |1 |0 |21611|320|50 |2480|3 |297 |100111|61 |0.0 |0.0 |0.0 |0.0 |4.0 |4.0 |1.0 |0.0 |23751.0 |64.0 |

|10000987464039884177|0 |14102816|1005|0 |5bcf81a2|9d54950b |f028772b |ecad2386|7801e8d9 |07d7df22 |a99f214a |845f69f4 |fa61e8fe |1 |0 |23438|320|50 |2684|2 |1319|-1 |52 |0.0 |11.0 |7.0 |1.0 |0.0 |0.0 |0.0 |0.0 |5237.0 |67.0 |

|10001055656394300907|0 |14102814|1005|0 |85f751fd|c4e18dd6 |50e219e0 |e9739828|df32afa9 |cef3e649 |a99f214a |6454c6ba |ecb851b2 |1 |0 |23441|320|50 |2685|1 |33 |100083|212|0.0 |0.0 |0.0 |0.0 |13.0 |11.0 |2.0 |0.0 |18147.0 |8.0 |

|10001237608243220141|0 |14102701|1005|0 |85f751fd|c4e18dd6 |50e219e0 |febd1138|82e27996 |0f2161f8 |a99f214a |ab986e15 |2ea4f8ba |1 |0 |19743|320|50 |2264|3 |427 |100000|61 |0.0 |0.0 |0.0 |0.0 |4.0 |4.0 |1.0 |0.0 |23941.0 |34.0 |

|10001363001408225332|0 |14102812|1005|1 |85f751fd|c4e18dd6 |50e219e0 |1dc72b4d|2347f47a |0f2161f8 |b7c2e4b6 |bce45090 |5db079b5 |1 |2 |22998|300|50 |2657|3 |35 |100013|23 |0.0 |0.0 |0.0 |0.0 |25.0 |1.0 |1.0 |1760.0 |729.0 |25.0 |

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+----+------+---+-----------+-------------+-----------------+-------------------+------------+----------------+------------------+---------------+---------------+------------------+

only showing top 5 rows

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+---+------+---+-----------+-------------+-----------------+-------------------+------------+----------------+------------------+---------------+---------------+------------------+

|id |click|hour |C1 |banner_pos|site_id |site_domain|site_category|app_id |app_domain|app_category|device_id|device_ip|device_model|device_type|device_conn_type|C14 |C15|C16|C17 |C18|C19|C20 |C21|click_index|site_id_index|site_domain_index|site_category_index|app_id_index|app_domain_index|app_category_index|device_id_index|device_ip_index|device_model_index|

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+---+------+---+-----------+-------------+-----------------+-------------------+------------+----------------+------------------+---------------+---------------+------------------+

|10002333262420133303|0 |14102211|1005|1 |856e6d3f|58a89a43 |f028772b |ecad2386|7801e8d9 |07d7df22 |a99f214a |ac322dfb |0dc22ebc |1 |0 |19771|320|50 |2227|0 |679|100077|48 |0.0 |6.0 |6.0 |1.0 |0.0 |0.0 |0.0 |0.0 |6004.0 |279.0 |

|10002749335348787004|1 |14102800|1005|0 |2a68aa20|9b851bd8 |3e814130 |ecad2386|7801e8d9 |07d7df22 |a99f214a |b4a0ec64 |49bc419a |1 |0 |20213|320|50 |2316|0 |167|100079|16 |1.0 |57.0 |56.0 |3.0 |0.0 |0.0 |0.0 |0.0 |30.0 |563.0 |

|10003763177308262205|0 |14102814|1002|0 |7971d583|c4e18dd6 |50e219e0 |ecad2386|7801e8d9 |07d7df22 |fffcf8a4 |f615f762 |a5df7413 |0 |0 |23441|320|50 |2685|1 |33 |-1 |212|0.0 |408.0 |0.0 |0.0 |0.0 |0.0 |0.0 |1003.0 |5471.0 |982.0 |

|10005435104591133943|0 |14102719|1005|0 |85f751fd|c4e18dd6 |50e219e0 |92f5800b|ae637522 |0f2161f8 |a99f214a |8f2784a2 |0bcabeaf |1 |3 |21189|320|50 |2424|1 |161|100193|71 |0.0 |0.0 |0.0 |0.0 |1.0 |2.0 |1.0 |0.0 |4207.0 |19.0 |

|10006076676750034840|0 |14102522|1005|1 |e151e245|7e091613 |f028772b |ecad2386|7801e8d9 |07d7df22 |a99f214a |dc88197f |fce66524 |1 |0 |4687 |320|50 |423 |2 |39 |100148|32 |0.0 |2.0 |2.0 |1.0 |0.0 |0.0 |0.0 |0.0 |4109.0 |22.0 |

+--------------------+-----+--------+----+----------+--------+-----------+-------------+--------+----------+------------+---------+---------+------------+-----------+----------------+-----+---+---+----+---+---+------+---+-----------+-------------+-----------------+-------------------+------------+----------------+------------------+---------------+---------------+------------------+

only showing top 5 rows

3. LR模型训练和预测

采用spark ml中LR模型，对广告点击进行预测。其中一些设置参数如下：

• setMaxIter设置最大迭代次数(默认100),具体迭代次数可能在不足最大迭代次数停止(见下一条)
• setTol设置容错(默认1e-6),每次迭代会计算一个误差,误差值随着迭代次数增加而减小,当误差小于设置容错,则停止迭代
• setRegParam设置正则化项系数(默认0),正则化主要用于防止过拟合现象,如果数据集较小,特征维数又多,易出现过拟合,考虑增大正则化系数
• setElasticNetParam正则化范式比(默认0),正则化有两种方式:L1(Lasso)和L2(Ridge),L1用于特征的稀疏化,L2用于防止过拟合
• setLabelCol设置标签列
• setFeaturesCol设置特征列
• setPredictionCol设置预测列
• setThreshold设置二分类阈值

val lr = new LogisticRegression()

.setMaxIter(10)

.setRegParam(0.3)

.setElasticNetParam(0)

.setFeaturesCol("feature")

.setLabelCol("click_index")

.setPredictionCol("click_predict")

val model_lr = lr.fit(train_hs)

println(s"每个特征对应系数: ${model_lr.coefficients} 截距: ${model_lr.intercept}")

val predictions = model_lr.transform(test_hs)

predictions.select("click_index","click_predict","probability").show(10,false)

val predictionRdd = predictions.select("click_predict","click_index").rdd.map{

case Row(click_predict:Double,click_index:Double)=>(click_predict,click_index)

}

val metrics = new MulticlassMetrics(predictionRdd)

val accuracy = metrics.accuracy

val weightedPrecision = metrics.weightedPrecision

val weightedRecall = metrics.weightedRecall

val f1 = metrics.weightedFMeasure

println(s"LR评估结果：\n分类正确率：${accuracy}\n加权正确率：${weightedPrecision}\n加权召回率：${weightedRecall}\nF1值：${f1}")

+-----------+-------------+----------------------------------------+

|click_index|click_predict|probability |

+-----------+-------------+----------------------------------------+

|0.0 |0.0 |[0.8673583515173942,0.13264164848260582]|

|1.0 |0.0 |[0.7065355297971061,0.29346447020289396]|

|0.0 |0.0 |[0.9247213791421071,0.07527862085789287]|

|0.0 |0.0 |[0.9411799267286762,0.05882007327132381]|

|0.0 |0.0 |[0.7534455683444734,0.24655443165552665]|

|0.0 |0.0 |[0.8993737856386326,0.10062621436136741]|

|0.0 |0.0 |[0.8837461636081269,0.11625383639187312]|

|0.0 |0.0 |[0.8320314092251319,0.16796859077486806]|

|0.0 |0.0 |[0.9027137639161569,0.09728623608384318]|

|1.0 |0.0 |[0.8791816482313737,0.12081835176862625]|

+-----------+-------------+----------------------------------------+

only showing top 10 rows

LR评估结果：

分类正确率：0.8308678500986193

加权正确率：0.7886992955593048

加权召回率：0.8308678500986193

F1值：0.7596712330402737

三、总结

1. 流程回顾

2.源码

object AdsCtrPredictionLR {
  def main(args: Array[String]): Unit = {
    val spark = SparkSession.builder()
      .appName("AdsCtrPredictionLR")
      .master("local[2]")
      .config("spark.some.config.option", "some-value")
      .getOrCreate()
     /**
      * id和click分别为广告的id和是否点击广告
      * site_id,site_domain,site_category,app_id,app_domain,app_category,device_id,device_ip,device_model为分类特征，需要OneHot编码
      * device_type,device_conn_type,C14,C15,C16,C17,C18,C19,C20,C21为数值特征，直接使用
      */
    val data = spark.read.csv("/opt/data/ads_6M.csv").toDF(
      "id","click","hour","C1","banner_pos","site_id","site_domain",
      "site_category","app_id","app_domain","app_category","device_id","device_ip",
      "device_model","device_type","device_conn_type","C14","C15","C16","C17","C18",
      "C19","C20","C21")
    data.show(5,false)
     val splited = data.randomSplit(Array(0.7,0.3),2L)
    val catalog_features = Array("click","site_id","site_domain","site_category","app_id","app_domain","app_category","device_id","device_ip","device_model")
    var train_index = splited(0)
    var test_index = splited(1)
    for(catalog_feature <- catalog_features){
      val indexer = new StringIndexer()
        .setInputCol(catalog_feature)
        .setOutputCol(catalog_feature.concat("_index"))
      val train_index_model = indexer.fit(train_index)
      val train_indexed = train_index_model.transform(train_index)
      val test_indexed = indexer.fit(test_index).transform(test_index,train_index_model.extractParamMap())
      train_index = train_indexed
      test_index = test_indexed
    }
    println("字符串编码下标标签：")
    train_index.show(5,false)
    test_index.show(5,false)
 //    特征Hasher
    val hasher = new FeatureHasher()
      .setInputCols("site_id_index","site_domain_index","site_category_index","app_id_index","app_domain_index","app_category_index","device_id_index","device_ip_index","device_model_index","device_type","device_conn_type","C14","C15","C16","C17","C18","C19","C20","C21")
      .setOutputCol("feature")
     println("特征Hasher编码：")
    val train_hs = hasher.transform(train_index)
    val test_hs = hasher.transform(test_index)
     /**
      * LR建模
      * setMaxIter设置最大迭代次数(默认100),具体迭代次数可能在不足最大迭代次数停止(见下一条)
      * setTol设置容错(默认1e-6),每次迭代会计算一个误差,误差值随着迭代次数增加而减小,当误差小于设置容错,则停止迭代
      * setRegParam设置正则化项系数(默认0),正则化主要用于防止过拟合现象,如果数据集较小,特征维数又多,易出现过拟合,考虑增大正则化系数
      * setElasticNetParam正则化范式比(默认0),正则化有两种方式:L1(Lasso)和L2(Ridge),L1用于特征的稀疏化,L2用于防止过拟合
      * setLabelCol设置标签列
      * setFeaturesCol设置特征列
      * setPredictionCol设置预测列
      * setThreshold设置二分类阈值
      */
    val lr = new LogisticRegression()
      .setMaxIter(10)
      .setRegParam(0.3)
      .setElasticNetParam(0)
      .setFeaturesCol("feature")
      .setLabelCol("click_index")
      .setPredictionCol("click_predict")
     val model_lr = lr.fit(train_hs)
     println(s"每个特征对应系数: ${model_lr.coefficients} 截距: ${model_lr.intercept}")
     val predictions = model_lr.transform(test_hs)
    predictions.select("click_index","click_predict","probability").show(100,false)
     val predictionRdd = predictions.select("click_predict","click_index").rdd.map{
      case Row(click_predict:Double,click_index:Double)=>(click_predict,click_index)
    }
    val metrics = new MulticlassMetrics(predictionRdd)
     val accuracy = metrics.accuracy
    val weightedPrecision = metrics.weightedPrecision
    val weightedRecall = metrics.weightedRecall
    val f1 = metrics.weightedFMeasure
     println(s"LR评估结果：\n分类正确率：${accuracy}\n加权正确率：${weightedPrecision}\n加权召回率：${weightedRecall}\nF1值：${f1}")
   }
}

 参考文献

https://blog.csdn.net/xueqingdata/article/details/50578005

https://blog.csdn.net/yhao2014/article/details/60324939

http://spark.apache.org/docs/latest/ml-features.html

http://spark.apache.org/docs/latest/ml-classification-regression.html