【机器学习笔记四】分类算法 - 逻辑回归

package com.fredric.spark.logistic

import org.apache.spark.mllib.classification.LogisticRegressionWithSGD
import org.apache.spark.mllib.linalg.Vectors
import org.apache.spark.mllib.regression.LabeledPoint
import org.apache.spark.{SparkContext, SparkConf}

/*-
 * 逻辑回归
 * Fredric 2017
 */
object logistic {
  def main(args:Array[String]): Unit ={

    val conf = new SparkConf().setMaster("local").setAppName("logistic")
    val sc   = new SparkContext(conf)

    val Array = new Array[LabeledPoint](10)

    //构造训练数据，虚拟一个以5为值的分类
    //针对一元二项逻辑斯特回归分类
    for(i <- 0 to 9){
      if(i >= 5){
        Array(i) = new LabeledPoint(1,Vectors.dense(i))
      }else{
        Array(i) = new LabeledPoint(0,Vectors.dense(i))
      }
    }

    val data = sc.makeRDD(Array);
    val model = LogisticRegressionWithSGD.train(data, 50)

    //model.weights输出[0.20670127500478114]
    println(model.weights)

    var test = -2

    //当输入为-1时，返回概率为0.0
    //当输入为11时，返回概率为1.0
    val result = model.predict(Vectors.dense(test))
    println(result)

    //验证该方式
    //计算P(Y=1|X)，测算输入x返回1的条件概率

    val res1 = Math.exp(model.weights(0)*test)/(1 + Math.exp(model.weights(0)*test))
    //计算P(Y=0|X)，测算输入x返回0的条件概率
    val res0 = 1/(1 + Math.exp(model.weights(0) * test))

    //输出：for target:-2 propalitity for 1 is : 0.3980965348017618 propalitity for 0 is : 0.6019034651982381
    //根据两个条件概率的比较可知-2属于分类0
    println("for target:" + test + " propalitity for 1 is : " + res1 + " propalitity for 0 is : " + res0)
  }
}