wzhe06/sparrowrecsys

A Deep Learning Recommender System

SparrowRecSys

SparrowRecSys是一个电影推荐系统，名字SparrowRecSys（麻雀推荐系统），取自“麻雀虽小，五脏俱全”之意。项目是一个基于maven的混合语言项目，同时包含了TensorFlow，Spark，Jetty Server等推荐系统的不同模块。希望你能够利用SparrowRecSys进行推荐系统的学习，并有机会一起完善它。

基于SparrowRecSys的实践课程

受极客时间邀请开设 深度学习推荐系统实战 课程，详细讲解了SparrowRecSys的所有技术细节，覆盖了深度学习模型结构，模型训练，特征工程，模型评估，模型线上服务及推荐服务器内部逻辑等模块。

环境要求

• Java 8
• Scala 2.11
• Python 3.6+
• TensorFlow 2.0+

快速开始

将项目用IntelliJ打开后，找到RecSysServer，右键点选Run，然后在浏览器中输入http://localhost:6010/即可看到推荐系统的前端效果。

项目数据

项目数据来源于开源电影数据集MovieLens，项目自带数据集对MovieLens数据集进行了精简，仅保留1000部电影和相关评论、用户数据。全量数据集请到MovieLens官方网站进行下载，推荐使用MovieLens 20M Dataset。

SparrowRecSys技术架构

SparrowRecSys技术架构遵循经典的工业级深度学习推荐系统架构，包括了离线数据处理、模型训练、近线的流处理、线上模型服务、前端推荐结果显示等多个模块。以下是SparrowRecSys的架构图：

SparrowRecSys实现的深度学习模型

• Word2vec (Item2vec)
• DeepWalk (Random Walk based Graph Embedding)
• Embedding MLP
• Wide&Deep
• Nerual CF
• Two Towers
• DeepFM
• DIN(Deep Interest Network)

相关论文

• [FFM] Field-aware Factorization Machines for CTR Prediction (Criteo 2016)
  
• [GBDT+LR] Practical Lessons from Predicting Clicks on Ads at Facebook (Facebook 2014)
  
• [PS-PLM] Learning Piece-wise Linear Models from Large Scale Data for Ad Click Prediction (Alibaba 2017)
  
• [FM] Fast Context-aware Recommendations with Factorization Machines (UKON 2011)
  
• [DCN] Deep & Cross Network for Ad Click Predictions (Stanford 2017)
  
• [Deep Crossing] Deep Crossing - Web-Scale Modeling without Manually Crafted Combinatorial Features (Microsoft 2016)
  
• [PNN] Product-based Neural Networks for User Response Prediction (SJTU 2016)
  
• [DIN] Deep Interest Network for Click-Through Rate Prediction (Alibaba 2018)
  
• [ESMM] Entire Space Multi-Task Model - An Effective Approach for Estimating Post-Click Conversion Rate (Alibaba 2018)
  
• [Wide & Deep] Wide & Deep Learning for Recommender Systems (Google 2016)
  
• [xDeepFM] xDeepFM - Combining Explicit and Implicit Feature Interactions for Recommender Systems (USTC 2018)
  
• [Image CTR] Image Matters - Visually modeling user behaviors using Advanced Model Server (Alibaba 2018)
  
• [AFM] Attentional Factorization Machines - Learning the Weight of Feature Interactions via Attention Networks (ZJU 2017)
  
• [DIEN] Deep Interest Evolution Network for Click-Through Rate Prediction (Alibaba 2019)
  
• [DSSM] Learning Deep Structured Semantic Models for Web Search using Clickthrough Data (UIUC 2013)
  
• [FNN] Deep Learning over Multi-field Categorical Data (UCL 2016)
  
• [DeepFM] A Factorization-Machine based Neural Network for CTR Prediction (HIT-Huawei 2017)
  
• [NFM] Neural Factorization Machines for Sparse Predictive Analytics (NUS 2017)
  

其他相关资源

• Papers on Computational Advertising
  
• Papers on Recommender System
  
• CTR Model Based on Spark
  

sparrowrecsys's Issues

推荐领域

基于这个项目，我是否可以改成文章推荐系统？

直接运行scala文件报错

请问这是什么问题

关于随机游走采样的方式

王老师您好，对于代码中randomwalk随机采样方式，有一些疑问，代码如下：

    while i < sampleLength:
        if (curElement not in itemDistribution) or (curElement not in transitionMatrix):
            break
        probDistribution = transitionMatrix[curElement]
        randomDouble = random.random()
        accumulateProb = 0.0
        for item, prob in probDistribution.items():
            accumulateProb += prob
            if accumulateProb >= randomDouble:
                curElement = item
                break
        sample.append(curElement)
        i += 1

对于上面的代码我的理解是先随机一个概率randomDouble然后按某个顺序probDistribution.items()遍历当前结点的所有邻接点，每个邻接点概率与原始概率accumulateProb相加，当累计相加大于randomDouble时就选择这个结点放入路径中。

个人感觉这种方式没有利用到probDistribution.items()中的概率分布，即便是使用无权边也即所有邻接点的概率相同，用老师的这种方式也是可以做的。我的想法是是否能够充分利用到邻接点的概率分布，对于概率大的（也即原始pair出现次数多的）给与更大的采样概率？ 使用np.random.choice(adj_nodes, node_distributions)可以实现这种采样方式，而且效率可能比老师这种循环的方式来得更快。

而按老师的这种采样方式，如果将probDistribution.items()先按从大到小排序，然后再采样，应该也能达到充分利用邻接点概率分布的效果。

不知道我的理解对不对？

DIEN中 auxiliary_loss_layer实现疑问

SparrowRecSys/TFRecModel/src/com/sparrowrecsys/offline/tensorflow/DIEN.py

Line 261 in 579577c

class auxiliary_loss_layer(tf.keras.layers.Layer):

请问一下auxiliary_loss_layer的实现逻辑是将positive samples 和 negative samples 分别通过两个不同的网络分支，这不会导致每个网络分支输入的样本都只有正例，没有负例，分布极不均衡，达不到训练目的吗？

graph embedding中transitionCountMatrix的计算问题

源代码：

def generateTransitionMatrix(samples):
    pairSamples = samples.flatMap(lambda x: generate_pair(x))
    pairCountMap = pairSamples.countByValue()
    pairTotalCount = 0
    transitionCountMatrix = defaultdict(dict)
    itemCountMap = defaultdict(int)
    for key, cnt in pairCountMap.items():
        key1, key2 = key

        # 此处是否应该改为 += cnt
        transitionCountMatrix[key1][key2] = cnt
        itemCountMap[key1] += cnt
        pairTotalCount += cnt
        ......

修改：

    for key, cnt in pairCountMap.items():
        key1, key2 = key

        if key1 not in transitionCountMatrix or key2 not in transitionCountMatrix[key1]:
            transitionCountMatrix[key1][key2] = cnt
        else:
            transitionCountMatrix[key1][key2] += cnt

        itemCountMap[key1] += cnt
        pairTotalCount += cnt

一些不合理的地方？

你好，我在将系统运行起来之后，发现User136的推荐列表中出现了他曾看过的电影，这是否合理？

embedding.scala

userEmb = user._2.foldRightArray[Float]((row, newEmb) => {
val movieId = row.getAsString
val movieEmb = word2VecModel.getVectors.get(movieId)
if(movieEmb.isDefined){
newEmb.zip(movieEmb.get).map { case (x, y) => x + y }
}else{
newEmb
}
})

老师，这段代码能解释下吗？user._2.foldRightArray[Float],(userEmb) 是什么意思，row, newEmb，这些参数是啥意思 ，movieEmb.get 为啥能这么写，是什么意思，
if(movieEmb.isDefined){
newEmb.zip(movieEmb.get).map { case (x, y) => x + y }
}else{
newEmb
}
这个又是起什么作用？

WARNING:tensorflow:Layers in a Sequential model should only have a single input tensor, but we receive a <class 'collections.OrderedDict'> input

在实现MLP的实例的时候，出现这个问题，这个是要怎么解决呢。

因为是个小白，请多多包涵。

谢谢。

运行jar包时提示：java.io.FileNotFoundException: nullsampledata/movies.csv

$ java -jar SparrowRecSys-1.0-SNAPSHOT-jar-with-dependencies.jar
log4j:WARN No appenders could be found for logger (org.eclipse.jetty.util.log).
log4j:WARN Please initialize the log4j system properly.
log4j:WARN See http://logging.apache.org/log4j/1.2/faq.html#noconfig for more info.
webRootLocation.toURI().toASCIIString(): jar:file:/mnt/e/work/SparrowRecSys/target/SparrowRecSys-1.0-SNAPSHOT-jar-with-dependencies.jar!/webroot/index.html
Web Root URI: null
Loading movie data from nullsampledata/movies.csv ...
Exception in thread "main" java.io.FileNotFoundException: nullsampledata/movies.csv (No such file or directory)
at java.base/java.io.FileInputStream.open0(Native Method)
at java.base/java.io.FileInputStream.open(FileInputStream.java:219)
at java.base/java.io.FileInputStream.(FileInputStream.java:157)
at java.base/java.util.Scanner.(Scanner.java:639)
at com.sparrowrecsys.online.datamanager.DataManager.loadMovieData(DataManager.java:56)
at com.sparrowrecsys.online.datamanager.DataManager.loadData(DataManager.java:41)
at com.sparrowrecsys.online.RecSysServer.run(RecSysServer.java:54)
at com.sparrowrecsys.online.RecSysServer.main(RecSysServer.java:21)

抄袭问题

https://pdf.hanspub.org/CSA20210700000_49286196.pdf

这个论文是抄袭了作者的项目吗？

Cannot resolve org.apache.htrace:htrace-core:3.1.0-incubating

Cannot resolve org.apache.htrace:htrace-core:3.1.0-incubating

DIEN模型的loss为负数

DIEN模型bug？我使用sampledata进行测试，loss一直都是负数，感觉不太对，不确定这是什么带来的

tensorflow运行结果
Epoch 1/5
/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/functional.py:591: UserWarning: Input dict contained keys ['rating', 'timestamp', 'userAvgReleaseYear', 'userReleaseYearStddev'] which did not match any model input. They will be ignored by the model.
[n for n in tensors.keys() if n not in ref_input_names])
7403/7403 [==============================] - 175s 23ms/step - loss: 3.4184 - auc_6: 0.5772 - auc_value: 0.5565
Epoch 2/5
7403/7403 [==============================] - 162s 22ms/step - loss: -3.2109 - auc_6: 0.6725 - auc_value: 0.6455
Epoch 3/5
7403/7403 [==============================] - 162s 22ms/step - loss: -3.4127 - auc_6: 0.7511 - auc_value: 0.7350
Epoch 4/5
7403/7403 [==============================] - 162s 22ms/step - loss: -3.4589 - auc_6: 0.7959 - auc_value: 0.7831
Epoch 5/5
7403/7403 [==============================] - 162s 22ms/step - loss: -3.4888 - auc_6: 0.8209 - auc_value: 0.8140
1870/1870 [==============================] - 23s 11ms/step - loss: -3.3402 - auc_6: 0.7502 - auc_value: 0.7512

Test Loss -3.3401615619659424, Test Accuracy 0.7501789331436157, Test ROC AUC 0.7511806488037109,

以及DIEN.py的304行报错，
test_loss, test_roc_auc = model.evaluate(test_dataset)
ValueError: too many values to unpack (expected 2)
感觉应该是
test_loss, test_roc_auc ,test_accuracy = model.evaluate(test_dataset)

请教下关于FM如何处理序列的问题

王喆老师，请问下，我想参考张俊林老师的博客推荐系统召回四模型之：全能的FM模型
做一个FM的embedding召回。如果我想把用户看过的电影序列输入FM中，每个电影是否应该同属于一个field呢？

我的理解是电影之间似乎不需要什么交互，因此用户看过的所有电影作为特征输入是应该标注成为一个field。

另一个就是电影的不同类别（gener）特征，也即动作，悬疑，科幻这种，也是序列，但是这种应该是可以有特征交互的，所以输入FM中的时候应该标注为不同Field。

不知这个理解是否正确？

运行时报错

java: 程序包org.apache.flink.api.common.functions不存在

请问该怎么解决？

ERR wrong number of arguments for 'hset' command

用 spark 跑那个 特征工程的时候，把结果往 redis 里面写的时候，报错了，错误信息如下：

ERR wrong number of arguments for 'hset' command

错误的代码是这行：
redisClient.hset(userKey, JavaConversions.mapAsJavaMap(valueMap))

向量文件写不进来

Embedding.scala 如下代码，写操作无用，为何呢？？？？？？？

val bw = new BufferedWriter(new FileWriter(file))
for (movieId <- model.getVectors.keys) {
  bw.write(movieId + "::" + model.getVectors(movieId).mkString(" ") + "\n")
  println(model.getVectors(movieId).mkString(" "))
}
bw.close()

执行FeatureEngineering 报错

Failed to execute goal org.codehaus.mojo:exec-maven-plugin:1.2:exec (default-cli) on project SparrowRecSys: Command execution failed.

DeepFM.py 运训报错

python:3.8.5
tensorflow:2.2.0
tensorflow.python.framework.errors_impl.UnimplementedError: Cast string to int32 is not supported

randomwalk函数

老师，你好，我是极客时间的学员，请问这个embedding.scala 中randomwalk函数 最后一行代码 Seq(sample.toList : _*)
sample.toList : 下划线*是什么意思？ 冒号及下划线*是干嘛的？

运行FeatureEngForRecModel.py时报错

21/02/23 10:54:55 ERROR TaskSetManager: Task 0 in stage 10.0 failed 1 times; aborting job
21/02/23 10:54:55 WARN TaskSetManager: Lost task 0.0 in stage 11.0 (TID 210, localhost, executor driver): TaskKilled (Stage cancelled)
Traceback (most recent call last):
File "D:/code/sparrowrecsys/SparrowRecSys-master/RecPySpark/src/com/sparrowrecsys/offline/pyspark/featureeng/FeatureEngForRecModel.py", line 151, in
samplesWithMovieFeatures = addMovieFeatures(movieSamples, ratingSamplesWithLabel)
File "D:/code/sparrowrecsys/SparrowRecSys-master/RecPySpark/src/com/sparrowrecsys/offline/pyspark/featureeng/FeatureEngForRecModel.py", line 54, in addMovieFeatures
samplesWithMovies4.show(5, truncate=False)
File "D:\ProgramData\Anaconda3\envs\recoenv\lib\site-packages\pyspark\sql\dataframe.py", line 380, in show
print(self._jdf.showString(n, int(truncate), vertical))
File "D:\ProgramData\Anaconda3\envs\recoenv\lib\site-packages\py4j\java_gateway.py", line 1257, in call
answer, self.gateway_client, self.target_id, self.name)
File "D:\ProgramData\Anaconda3\envs\recoenv\lib\site-packages\pyspark\sql\utils.py", line 63, in deco
return f(*a, **kw)
File "D:\ProgramData\Anaconda3\envs\recoenv\lib\site-packages\py4j\protocol.py", line 328, in get_return_value
format(target_id, ".", name), value)
py4j.protocol.Py4JJavaError: An error occurred while calling o132.showString.
: org.apache.spark.SparkException: Job aborted due to stage failure: Task 0 in stage 10.0 failed 1 times, most recent failure: Lost task 0.0 in stage 10.0 (TID 209, localhost, executor driver): org.apache.spark.SparkException: Python worker failed to connect back.

建议API使用SpringBoot

个人建议API使SpringBoot， 如此代码更简单明了，而且熟悉的人也多。 当然Servlet是基础，如果需要可以，我可以帮忙

Plugin 'org.apache.maven.plugins:maven-compiler-plugin:' not found

建议上传一份数据，方便小白（我）玩一下

# Training samples path, change to your local path
training_samples_file_path = tf.keras.utils.get_file("trainingSamples.csv",
                                                     "file:///Users/zhewang/Workspace/SparrowRecSys/src/main"
                                                     "/resources/webroot/sampledata/trainingSamples.csv")
# Test samples path, change to your local path
test_samples_file_path = tf.keras.utils.get_file("testSamples.csv",
                                                 "file:///Users/zhewang/Workspace/SparrowRecSys/src/main"
                                                 "/resources/webroot/sampledata/testSamples.csv")

主页上面的的搜索框为什么失效呀，我输入什么内容搜索完都还是跳转回主页

打开的推荐系统好像不太一样

Format with google-java-format

The code format is a little bit off. Will you consider formatting the code base with https://github.com/google/google-java-format, and furthermore make it part of the build process? This way no code without being properly formatted can be checked in.

tensorflow/serving上线问题

`[root@localhost temp]# docker run -t --rm -p 8501:8501 \

-v "$TESTDATA/saved_model_half_plus_two_cpu:/models/half_plus_two" \
-e MODEL_NAME=half_plus_two \
tensorflow/serving &

[1] 2053
[root@localhost temp]# /usr/bin/tf_serving_entrypoint.sh: line 3: 7 Illegal instruction (core dumped) tensorflow_model_server --port=8500 --rest_api_port=8501 --model_name=${MODEL_NAME} --model_base_path=${MODEL_BASE_PATH}/${MODEL_NAME} "$@"

[1]+ Exit 132 docker run -t --rm -p 8501:8501 -v "$TESTDATA/saved_model_half_plus_two_cpu:/models/half_plus_two" -e MODEL_NAME=half_plus_two tensorflow/serving
[root@localhost temp]# `

在docker里面【我的docker是安装在centos7中】执行上面那个命令的时候，一直都是出现上面这种情况，然后模型无法上线，请问有什么好的办法吗？

个性化推荐代码在哪里啊?

已经实现了吗？

运行FeatureEngForRecModel.py报错

@bfgf52 您好，我在jupyter中运行这个文件中的代码
运行到这一步时报错：samplesWithUserFeatures = addUserFeatures(samplesWithMovieFeatures)
报错如下，麻烦帮忙看看问题出在哪？谢谢：
Py4JJavaError: An error occurred while calling o233.withColumn.
: org.apache.spark.sql.AnalysisException: cannot resolve 'reverse(userPositiveHistory)' due to data type mismatch: argument 1 requires string type, however, 'userPositiveHistory' is of array type.;;
'Project [movieId#27, userId#26, rating#28, timestamp#29, label#88, genres#12, releaseYear#122, movieGenre1#147, movieGenre2#156, movieGenre3#166, movieRatingCount#233L, movieAvgRating#190, movieRatingStddev#239, reverse(userPositiveHistory#354) AS userPositiveHistory#370]
+- Project [movieId#27, userId#26, rating#28, timestamp#29, label#88, genres#12, releaseYear#122, movieGenre1#147, movieGenre2#156, movieGenre3#166, movieRatingCount#233L, movieAvgRating#190, movieRatingStddev#239, userPositiveHistory#354]
+- Project [movieId#27, userId#26, rating#28, timestamp#29, label#88, genres#12, releaseYear#122, movieGenre1#147, movieGenre2#156, movieGenre3#166, movieRatingCount#233L, movieAvgRating#190, movieRatingStddev#239, _w0#355, userPositiveHistory#354, userPositiveHistory#354]
+- Window [collect_list(_w0#355, 0, 0) windowspecdefinition(userId#26, timestamp#29 ASC NULLS FIRST, specifiedwindowframe(RowFrame, -100, -1)) AS userPositiveHistory#354], [userId#26], [timestamp#29 ASC NULLS FIRST]
+- Project [movieId#27, userId#26, rating#28, timestamp#29, label#88, genres#12, releaseYear#122, movieGenre1#147, movieGenre2#156, movieGenre3#166, movieRatingCount#233L, movieAvgRating#190, movieRatingStddev#239, CASE WHEN (label#88 = 1) THEN movieId#27 ELSE cast(null as string) END AS _w0#355]
+- Project [movieId#27, userId#26, rating#28, timestamp#29, label#88, genres#12, releaseYear#122, movieGenre1#147, movieGenre2#156, movieGenre3#166, movieRatingCount#233L, movieAvgRating#190, movieRatingStddev#239]
+- Join LeftOuter, (movieId#27 = movieId#245)
:- Project [movieId#27, userId#26, rating#28, timestamp#29, label#88, genres#12, releaseYear#122, movieGenre1#147, movieGenre2#156, split(genres#12, |)[2] AS movieGenre3#166]
: +- Project [movieId#27, userId#26, rating#28, timestamp#29, label#88, genres#12, releaseYear#122, movieGenre1#147, split(genres#12, |)[1] AS movieGenre2#156]
: +- Project [movieId#27, userId#26, rating#28, timestamp#29, label#88, genres#12, releaseYear#122, split(genres#12, |)[0] AS movieGenre1#147]
: +- Project [movieId#27, userId#26, rating#28, timestamp#29, label#88, genres#12, releaseYear#122]
: +- Project [movieId#27, userId#26, rating#28, timestamp#29, label#88, (title#11) AS title#131, genres#12, releaseYear#122]
: +- Project [movieId#27, userId#26, rating#28, timestamp#29, label#88, title#11, genres#12, extractReleaseYearUdf(title#11) AS releaseYear#122]
: +- Project [movieId#27, userId#26, rating#28, timestamp#29, label#88, title#11, genres#12]
: +- Join LeftOuter, (movieId#27 = movieId#10)
: :- Project [userId#26, movieId#27, rating#28, timestamp#29, CASE WHEN (cast(rating#28 as double) >= 3.5) THEN 1 ELSE 0 END AS label#88]
: : +- Relation[userId#26,movieId#27,rating#28,timestamp#29] csv
: +- Relation[movieId#10,title#11,genres#12] csv
+- Project [movieId#245, movieRatingCount#233L, movieAvgRating#190, format_number(movieRatingStddev#234, 2) AS movieRatingStddev#239]
+- Project [movieId#245, coalesce(movieRatingCount#188L, cast(0.0 as bigint)) AS movieRatingCount#233L, movieAvgRating#190, coalesce(nanvl(movieRatingStddev#200, cast(null as double)), cast(0.0 as double)) AS movieRatingStddev#234]
+- Aggregate [movieId#245], [movieId#245, count(1) AS movieRatingCount#188L, format_number(avg(cast(rating#246 as double)), 2) AS movieAvgRating#190, stddev_samp(cast(rating#246 as double)) AS movieRatingStddev#200]
+- Project [movieId#245, userId#244, rating#246, timestamp#247, label#88, genres#12, releaseYear#122, movieGenre1#147, movieGenre2#156, split(genres#12, |)[2] AS movieGenre3#166]
+- Project [movieId#245, userId#244, rating#246, timestamp#247, label#88, genres#12, releaseYear#122, movieGenre1#147, split(genres#12, |)[1] AS movieGenre2#156]
+- Project [movieId#245, userId#244, rating#246, timestamp#247, label#88, genres#12, releaseYear#122, split(genres#12, |)[0] AS movieGenre1#147]
+- Project [movieId#245, userId#244, rating#246, timestamp#247, label#88, genres#12, releaseYear#122]
+- Project [movieId#245, userId#244, rating#246, timestamp#247, label#88, (title#11) AS title#131, genres#12, releaseYear#122]
+- Project [movieId#245, userId#244, rating#246, timestamp#247, label#88, title#11, genres#12, extractReleaseYearUdf(title#11) AS releaseYear#122]
+- Project [movieId#245, userId#244, rating#246, timestamp#247, label#88, title#11, genres#12]
+- Join LeftOuter, (movieId#245 = movieId#10)
:- Project [userId#244, movieId#245, rating#246, timestamp#247, CASE WHEN (cast(rating#246 as double) >= 3.5) THEN 1 ELSE 0 END AS label#88]
: +- Relation[userId#244,movieId#245,rating#246,timestamp#247] csv
+- Relation[movieId#10,title#11,genres#12] csv

提示程序包不存在

C:\Users\xu\Downloads\SparrowRecSys-master\SparrowRecSys-master\src\main\java\com\sparrowrecsys\nearline\flink\RealTimeFeature.java:3:45
java: 程序包org.apache.flink.api.common.functions不存在
C:\Users\xu\Downloads\SparrowRecSys-master\SparrowRecSys-master\src\main\java\com\sparrowrecsys\nearline\flink\RealTimeFeature.java:4:36
java: 程序包org.apache.flink.api.java.io不存在
C:\Users\xu\Downloads\SparrowRecSys-master\SparrowRecSys-master\src\main\java\com\sparrowrecsys\nearline\flink\RealTimeFeature.java:5:49
java: 程序包org.apache.flink.streaming.api.datastream不存在
C:\Users\xu\Downloads\SparrowRecSys-master\SparrowRecSys-master\src\main\java\com\sparrowrecsys\nearline\flink\RealTimeFeature.java:6:50
java: 程序包org.apache.flink.streaming.api.environment不存在
C:\Users\xu\Downloads\SparrowRecSys-master\SparrowRecSys-master\src\main\java\com\sparrowrecsys\nearline\flink\RealTimeFeature.java:7:53
java: 程序包org.apache.flink.streaming.api.functions.sink不存在
C:\Users\xu\Downloads\SparrowRecSys-master\SparrowRecSys-master\src\main\java\com\sparrowrecsys\nearline\flink\RealTimeFeature.java:8:55
java: 程序包org.apache.flink.streaming.api.functions.source不存在
C:\Users\xu\Downloads\SparrowRecSys-master\SparrowRecSys-master\src\main\java\com\sparrowrecsys\nearline\flink\RealTimeFeature.java:9:53
java: 程序包org.apache.flink.streaming.api.windowing.time不存在
C:\Users\xu\Downloads\SparrowRecSys-master\SparrowRecSys-master\src\main\java\com\sparrowrecsys\nearline\flink\RealTimeFeature.java:34:15
java: 找不到符号
符号: 类 StreamExecutionEnvironment
位置: 类 com.sparrowrecsys.nearline.flink.RealTimeFeature
C:\Users\xu\Downloads\SparrowRecSys-master\SparrowRecSys-master\src\main\java\com\sparrowrecsys\nearline\flink\RealTimeFeature.java:34:48
java: 找不到符号
符号: 变量 StreamExecutionEnvironment
位置: 类 com.sparrowrecsys.nearline.flink.RealTimeFeature

项目完善程度

请问老师现在是一个完整项目了吗?感觉似乎很多模块不知道去哪里访问?比如召回排序等

如何设置用户推荐模型为“"nerualcf"”

将user.html中第136行左右的model设置为model = "nerualcf";后，在对用户进行推荐的时候，有的用户使用的模型仍然是“emb”，比如user170。这是因为什么？

Unimplemented: Cast string to int32 is not supported

when run WideNDeep.py,the bugs happens