12.6 地图实时路况预测

￡神魔★判官ぃ 2022-05-15 11:36 255阅读 0赞

PredictLRwithLBFGS.scala：

<table> 
 <tbody> 
 <tr> 
 <td> /**训练-逻辑回归 &nbsp;* Created by Administrator on 2018/8/24. &nbsp;*/ object TrainLRwithLBFGS { &nbsp; &nbsp; &nbsp; val sparkConf = new SparkConf().setAppName("Beijing traffic").setMaster("local") &nbsp; &nbsp; val sc = new SparkContext(sparkConf) &nbsp; &nbsp; &nbsp; // create the date/time formatters &nbsp; &nbsp; val dayFormat = new SimpleDateFormat("yyyyMMdd") &nbsp; &nbsp; val minuteFormat = new SimpleDateFormat("HHmm") &nbsp; &nbsp; &nbsp; def main(args: Array[String]) { &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // fetch data from redis &nbsp; &nbsp; &nbsp; &nbsp; val jedis = RedisClient.pool.getResource &nbsp; &nbsp; &nbsp; &nbsp; jedis.select(1) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // find relative road monitors for specified road &nbsp; &nbsp; &nbsp; &nbsp; // val camera_ids = List("310999003001","310999003102","310999000106","310999000205","310999007204") &nbsp; &nbsp; &nbsp; &nbsp; val camera_ids = List("310999003001","310999003102") &nbsp; &nbsp; &nbsp; &nbsp; val camera_relations:Map[String,Array[String]] = Map[String,Array[String]]( &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; "310999003001" -&gt; Array("310999003001","310999003102","310999000106","310999000205","310999007204"), &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; "310999003102" -&gt; Array("310999003001","310999003102","310999000106","310999000205","310999007204") &nbsp; &nbsp; &nbsp; &nbsp; ) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val temp = camera_ids.map({ camera_id =&gt; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val hours = 3 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val nowtimelong = System.currentTimeMillis(); &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val now = new Date(nowtimelong) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val day = dayFormat.format(now) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // Option Some None &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val list = camera_relations.get(camera_id).get &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val relations = list.map({ camera_id =&gt; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; println(camera_id) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // fetch records of one camera for three hours ago &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; (camera_id, jedis.hgetAll(day + "_" + camera_id)) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; }) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; relations.foreach(println) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // organize above records per minute to train data set format (MLUtils.loadLibSVMFile) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val trainSet = ArrayBuffer[LabeledPoint]() &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // start begin at index 3 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; //往训练集里面加上每一行数据，往前三小时，每次分钟数减1 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; for(i &lt;- Range(60*hours-3,0,-1)){ &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val featuresX = ArrayBuffer[Double]() &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val featureY = ArrayBuffer[Double]() &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // get current minute and recent two minutes &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; //填三分钟的数据 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; for(index &lt;- 0 to 2){ &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val tempOne = nowtimelong - 60 * 1000 * (i-index) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val d = new Date(tempOne) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val tempMinute = minuteFormat.format(d) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val tempNext = tempOne - 60 * 1000 * (-1) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val dNext = new Date(tempNext) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val tempMinuteNext = minuteFormat.format(dNext) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; //把自己和相邻路段填入值 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; for((k,v) &lt;- relations){ &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // k-&gt;camera_id ; v-&gt;HashMap &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val map = v &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; //填入y坐标值 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; if(index == 2 &amp;&amp; k == camera_id){ &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; if (map.containsKey(tempMinuteNext)) { &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val info = map.get(tempMinuteNext).split("_") &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val f = info(0).toFloat / info(1).toFloat &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; featureY += f &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; //填入X坐标值 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; if (map.containsKey(tempMinute)){ &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val info = map.get(tempMinute).split("_") &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val f = info(0).toFloat / info(1).toFloat &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; featuresX += f &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } else{ &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; featuresX += -1.0 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; //判断y坐标有没有值 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; if(featureY.toArray.length == 1 ){ &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val label = (featureY.toArray).head &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val record = LabeledPoint(if ((label.toInt/10)&lt;10) (label.toInt/10) else 10.0, Vectors.dense(featuresX.toArray)) // println(record) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; trainSet += record &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; trainSet.foreach(println) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; println(trainSet.length) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val data = sc.parallelize(trainSet) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; println(data) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // Split data into training (60%) and test (40%). &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val splits = data.randomSplit(Array(0.6, 0.4), seed = 1000L) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val training = splits(0) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val test = splits(1) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; if(!data.isEmpty()){ &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // Run training algorithm to build the model &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; //按逻辑回归模型训练--算法调用 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val model = new LogisticRegressionWithLBFGS() &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; .setNumClasses(11) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; .run(training) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // Compute raw scores on the test set. &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; //测试集测试 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val predictionAndLabels = test.map { case LabeledPoint(label, features) =&gt; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val prediction = model.predict(features) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; (prediction, label) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; predictionAndLabels.foreach(x=&gt; println(x)) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // Get evaluation metrics. &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; //结果评估 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val metrics = new MulticlassMetrics(predictionAndLabels) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val precision = metrics.precision &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; println("Precision = " + precision) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; //如果准确度大于80%，保存起来 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; if(precision &gt; 0.8){ &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val path = "hdfs://node1:9000/model_"+camera_id+"_"+nowtimelong &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; model.save(sc, path) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; println("saved model to "+ path) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; jedis.hset("model", camera_id , path) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp; &nbsp; &nbsp; }) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; RedisClient.pool.returnResource(jedis) &nbsp; &nbsp; } } </td> 
 </tr> 
 </tbody> 
</table>

PredictLRwithLBFGS.scala：

<table> 
 <tbody> 
 <tr> 
 <td> /**加载存储的模型，进行预测 &nbsp; &nbsp; * Created by Administrator on 2018/8/24. &nbsp; &nbsp; */ &nbsp; object PredictLRwithLBFGS { &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val sparkConf = new SparkConf().setAppName("Shanghai traffic").setMaster("local[4]") &nbsp; &nbsp; &nbsp; val sc = new SparkContext(sparkConf) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // create the date/time formatters &nbsp; &nbsp; &nbsp; val dayFormat = new SimpleDateFormat("yyyyMMdd") &nbsp; &nbsp; &nbsp; val minuteFormat = new SimpleDateFormat("HHmm") &nbsp; &nbsp; &nbsp; val sdf = new SimpleDateFormat( "yyyy-MM-dd_HH:mm:ss" ); &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; def main(args: Array[String]) { &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val input = "2017-04-20_09:50:00" &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val date = sdf.parse( input ); &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val inputTimeLong = date.getTime() &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val inputTime = new Date(inputTimeLong) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val day = dayFormat.format(inputTime) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // fetch data from redis &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val jedis = RedisClient.pool.getResource &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; jedis.select(1) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // find relative road monitors for specified road &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // val camera_ids = List("310999003001","310999003102","310999000106","310999000205","310999007204") &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val camera_ids = List("310999003001","310999003102") &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val camera_relations:Map[String,Array[String]] = Map[String,Array[String]]( &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; "310999003001" -&gt; Array("310999003001","310999003102","310999000106","310999000205","310999007204"), &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; "310999003102" -&gt; Array("310999003001","310999003102","310999000106","310999000205","310999007204") &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; ) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val temp = camera_ids.map({ camera_id =&gt; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val list = camera_relations.get(camera_id).get &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val relations = list.map({ camera_id =&gt; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; println(camera_id) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // fetch records of one camera for three hours ago &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; (camera_id, jedis.hgetAll(day + "_" + camera_id)) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; }) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; relations.foreach(println) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // organize above records per minute to train data set format (MLUtils.loadLibSVMFile) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val aaa = ArrayBuffer[Double]() &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // get current minute and recent two minutes &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; for(index &lt;- 0 to 2){ &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val tempOne = inputTimeLong - 60 * 1000 * index &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val tempMinute = minuteFormat.format(inputTime) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; for((k,v) &lt;- relations){ &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // k-&gt;camera_id ; v-&gt;speed &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val map = v &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; if (map.containsKey(tempMinute)){ &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val info = map.get(tempMinute).split("_") &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val f = info(0).toFloat / info(1).toFloat &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; aaa += f &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } else{ &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; aaa += -1.0 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // Run training algorithm to build the model &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; //读取训练好的模型，进行预测 &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val path = jedis.hget("model",camera_id) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val model = LogisticRegressionModel.load(sc, path) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; // Compute raw scores on the test set. &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; val prediction = model.predict(Vectors.dense(aaa.toArray)) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; println(input+"\t"+camera_id+"\t"+prediction+"\t") &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; jedis.hset(input, camera_id, prediction.toString) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; }) &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; RedisClient.pool.returnResource(jedis) &nbsp; &nbsp; &nbsp; } &nbsp; &nbsp;} </td> 
 </tr> 
 </tbody> 
</table>