大數(shù)據(jù)無處不在，推薦系統(tǒng)無處不在帆谍。
QQ音樂的今日推薦歌曲伪朽；人人網(wǎng)的好友推薦；新浪微博的你可能感覺興趣的人汛蝙；優(yōu)酷烈涮，土豆的電影推薦朴肺；豆瓣的圖書推薦；大從點(diǎn)評(píng)的餐飲推薦坚洽；世紀(jì)佳緣的相親推薦戈稿；天際網(wǎng)的職業(yè)推薦等都用到了大數(shù)據(jù)。
今天利用MapReduce簡(jiǎn)單寫個(gè)仿QQ音樂的推薦系統(tǒng)讶舰，希望能給在座各位在工作中或面試中一點(diǎn)幫助鞍盗！轉(zhuǎn)載請(qǐng)注明出處：Michael孟良
今日推薦歌曲
原理：
通過歷史對(duì)歌曲操作記錄，計(jì)算得出每首歌相對(duì)其他歌曲同時(shí)出現(xiàn)在同一用戶的次數(shù)跳昼，每件歌曲都有自己相對(duì)全部歌曲的同現(xiàn)列表般甲，用戶會(huì)對(duì)部分歌曲有過點(diǎn)擊，收藏等實(shí)際操作鹅颊，經(jīng)過計(jì)算會(huì)得到用戶對(duì)這部分歌曲的評(píng)分向量列表敷存。
使用用戶評(píng)分向量列表中的分值：
依次乘以每首歌同現(xiàn)列表中該分值的代表歌曲的同現(xiàn)值
求和便是該歌曲的推薦向量
具體算法：
推薦系統(tǒng)——協(xié)同過濾(Collaborative Filtering)算法
UserCF
基于用戶的協(xié)同過濾，通過不同用戶對(duì)歌曲的評(píng)分來評(píng)測(cè)用戶之間的相似性堪伍，基于用戶之間的相似性做出推薦锚烦。簡(jiǎn)單來講就是：給用戶推薦和他興趣相似的其他用戶喜歡的歌曲。
基于用戶的協(xié)同過濾
基于用戶的協(xié)同過濾
-ItemCF
基于item的協(xié)同過濾帝雇，通過用戶對(duì)不同item的評(píng)分來評(píng)測(cè)item之間的相似性涮俄，基于item之間的相似性做出推薦。簡(jiǎn)單來講就是：給用戶推薦和他之前喜歡的歌曲相似的歌曲尸闸。
基于item的協(xié)同過濾
基于item的協(xié)同過濾
—Co-occurrence Matrix(同現(xiàn)矩陣)和User Preference Vector(用戶評(píng)分向量)相乘得到的這個(gè)Recommended Vector(推薦向量)
—基于全量數(shù)據(jù)的統(tǒng)計(jì)彻亲，產(chǎn)生同現(xiàn)矩陣
1.體現(xiàn)歌曲間的關(guān)聯(lián)性
2.每首歌都有自己對(duì)其他全部歌曲的關(guān)聯(lián)性（每件歌曲的特征）
用戶評(píng)分向量體現(xiàn)的是用戶對(duì)一些歌曲的評(píng)分
任一歌曲需要：
1.用戶評(píng)分向量乘以基于該歌曲的其他歌曲關(guān)聯(lián)值
2.求和得出針對(duì)該歌曲的推薦向量
3.排序取TopN即可
公式
同現(xiàn)矩陣乘以用戶評(píng)分
好了，邏輯知道了室叉，怎么把它搬到代碼中睹栖，我們一步一步來：
數(shù)據(jù)：
item_id,user_id,action,vtime
i161,u2625,click,2018/8/18 15:03
i161,u2626,click,2018/8/23 22:40
i161,u2627,click,2018/8/25 19:09
i161,u2628,click,2018/8/28 21:35
i161,u2629,click,2018/8/27 16:33
i161,u2630,click,2018/8/5 18:45
i161,u2631,click,2018/8/29 16:57
i161,u2632,click,2018/8/24 21:58
i161,u2633,click,2018/8/25 22:41
i161,u2634,click,2018/8/16 13:30
i161,u2635,click,2018/8/20 9:23
i161,u2636,click,2018/8/21 1:00
i161,u2637,click,2018/8/24 22:51
...
...
...
歌曲id：item_id
用戶id：user_id
對(duì)歌曲操作：action
操作時(shí)間：vtime
代碼：
啟動(dòng)類StartRun
public class StartRun {
public static void main(String[] args) {
Configuration config = new Configuration(true);
config.set("mapreduce.framework.name", "local");
config.set("mapreduce.app-submission.cross-platform", "true");
// config.set("fs.defaultFS", "hdfs://node1:8020");
// config.set("yarn.resourcemanager.hostname", "node3");
// 所有mr的輸入和輸出目錄定義在map集合中
Map<String, String> paths = new HashMap<String, String>();
paths.put("Step1Input", "/user/root/m_log");
paths.put("Step1Output", "/data/itemcf/output/step1");
paths.put("Step2Input", paths.get("Step1Output"));
paths.put("Step2Output", "/data/itemcf/output/step2");
paths.put("Step3Input", paths.get("Step2Output"));
paths.put("Step3Output", "/data/itemcf/output/step3");
paths.put("Step4Input1", paths.get("Step2Output"));
paths.put("Step4Input2", paths.get("Step3Output"));
paths.put("Step4Output", "/data/itemcf/output/step4");
paths.put("Step5Input", paths.get("Step4Output"));
paths.put("Step5Output", "/data/itemcf/output/step5");
paths.put("Step6Input", paths.get("Step5Output"));
paths.put("Step6Output", "/data/itemcf/output/step6");
Step1.run(config, paths);
// Step2.run(config, paths);
// Step3.run(config, paths);
// Step4.run(config, paths);
// Step5.run(config, paths);
// Step6.run(config, paths);
}
public static Map<String, Integer> R = new HashMap<String, Integer>();
static {
R.put("click", 1);
R.put("share", 2);
R.put("like", 3);
R.put("download", 4);
}
}
我們分為6步：
第一步：
public class Step1 {
public static boolean run(Configuration config, Map<String, String> paths) {
try {
FileSystem fs = FileSystem.get(config);
Job job = Job.getInstance(config);
job.setJobName("step1");
// config.set("mapred.jar", "D:\\MR\\item.jar");
job.setJarByClass(Step1.class);
job.setMapperClass(Step1_Mapper.class);
job.setReducerClass(Step1_Reducer.class);
job.setMapOutputKeyClass(Text.class);
job.setMapOutputValueClass(NullWritable.class);
FileInputFormat.addInputPath(job, new Path(paths.get("Step1Input")));
Path outpath = new Path(paths.get("Step1Output"));
if (fs.exists(outpath)) {
fs.delete(outpath, true);
}
FileOutputFormat.setOutputPath(job, outpath);
boolean f = job.waitForCompletion(true);
return f;
} catch (Exception e) {
e.printStackTrace();
}
return false;
}
static class Step1_Mapper extends Mapper<LongWritable, Text, Text, NullWritable> {
@Override
protected void map(LongWritable key, Text value, Context context)
throws IOException, InterruptedException {
if (key.get() != 0) {
context.write(value, NullWritable.get());
}
}
}
static class Step1_Reducer extends Reducer<Text, IntWritable, Text, NullWritable> {
@Override
protected void reduce(Text key, Iterable<IntWritable> i, Context context)
throws IOException, InterruptedException {
context.write(key, NullWritable.get());
}
}
}
第一步去重硫惕，一進(jìn)一出茧痕，沒什么好說的。
第二步：
public class Step2 {
public static boolean run(Configuration config, Map<String, String> paths) {
try {
FileSystem fs = FileSystem.get(config);
Job job = Job.getInstance(config);
job.setJobName("step2");
job.setJarByClass(StartRun.class);
job.setMapperClass(Step2_Mapper.class);
job.setReducerClass(Step2_Reducer.class);
job.setMapOutputKeyClass(Text.class);
job.setMapOutputValueClass(Text.class);
FileInputFormat.addInputPath(job, new Path(paths.get("Step2Input")));
Path outpath = new Path(paths.get("Step2Output"));
if (fs.exists(outpath)) {
fs.delete(outpath, true);
}
FileOutputFormat.setOutputPath(job, outpath);
boolean f = job.waitForCompletion(true);
return f;
} catch (Exception e) {
e.printStackTrace();
}
return false;
}
static class Step2_Mapper extends Mapper<LongWritable, Text, Text, Text> {
// 如果使用：用戶+歌曲的id恼除，同時(shí)作為輸出key踪旷，更好
@Override
protected void map(LongWritable key, Text value, Context context)
throws IOException, InterruptedException {
//i1,u2723,click,2014/9/14 9:31
String[] tokens = value.toString().split(",");
String item = tokens[0];
String user = tokens[1];
String action = tokens[2];
Text k = new Text(user);
Integer rv = StartRun.R.get(action);
// if(rv!=null){
Text v = new Text(item + ":" + rv.intValue());
//u2750 i160:1
context.write(k, v);
}
}
static class Step2_Reducer extends Reducer<Text, Text, Text, Text> {
@Override
protected void reduce(Text key, Iterable<Text> i, Context context)
throws IOException, InterruptedException {
Map<String, Integer> r = new HashMap<String, Integer>();
//迭代同一用戶關(guān)注的歌曲的id
for (Text value : i) {
//u2750 i160:1
String[] vs = value.toString().split(":");
String item = vs[0];
Integer action = Integer.parseInt(vs[1]);
action = ((Integer) (r.get(item) == null ? 0 : r.get(item))).intValue() + action;
r.put(item, action);
}
StringBuffer sb = new StringBuffer();
for (Entry<String, Integer> entry : r.entrySet()) {
sb.append(entry.getKey() + ":" + entry.getValue().intValue() + ",");
}
//u2756 i105:1,i79:1,i341:1,i319:1,i332:1,i160:1,i342:1,i94:1,
context.write(key, new Text(sb.toString()));
}
}
}
第二步，按用戶分組豁辉，計(jì)算所有歌曲的id出現(xiàn)的組合列表令野，得到用戶對(duì)歌曲的id的喜愛度得分矩陣。
運(yùn)行結(jié)果：
第二步
第三步：
public class Step3 {
private final static Text K = new Text();
private final static IntWritable V = new IntWritable(1);
public static boolean run(Configuration config, Map<String, String> paths) {
try {
FileSystem fs = FileSystem.get(config);
Job job = Job.getInstance(config);
job.setJobName("step3");
job.setJarByClass(StartRun.class);
job.setMapperClass(Step3_Mapper.class);
job.setReducerClass(Step3_Reducer.class);
job.setCombinerClass(Step3_Reducer.class);
job.setMapOutputKeyClass(Text.class);
job.setMapOutputValueClass(IntWritable.class);
FileInputFormat
.addInputPath(job, new Path(paths.get("Step3Input")));
Path outpath = new Path(paths.get("Step3Output"));
if (fs.exists(outpath)) {
fs.delete(outpath, true);
}
FileOutputFormat.setOutputPath(job, outpath);
boolean f = job.waitForCompletion(true);
return f;
} catch (Exception e) {
e.printStackTrace();
}
return false;
}
// 第二個(gè)MR執(zhí)行的結(jié)果--作為本次MR的輸入 樣本: u2837 i541:1,i331:1,i314:1,i125:1,
static class Step3_Mapper extends
Mapper<LongWritable, Text, Text, IntWritable> {
@Override
protected void map(LongWritable key, Text value, Context context)
throws IOException, InterruptedException {
String[] tokens = value.toString().split("\t");
String[] items = tokens[1].split(",");
//嵌套循環(huán)徽级，每一個(gè)歌曲與其他歌曲組合輸出一次气破，val的值為1
//WC的思維邏輯
for (int i = 0; i < items.length; i++) {
String itemA = items[i].split(":")[0];
for (int j = 0; j < items.length; j++) {
String itemB = items[j].split(":")[0];
K.set(itemA + ":" + itemB);
context.write(K, V);
}
}
}
}
static class Step3_Reducer extends
Reducer<Text, IntWritable, Text, IntWritable> {
@Override
protected void reduce(Text key, Iterable<IntWritable> i, Context context)
throws IOException, InterruptedException {
int sum = 0;
for (IntWritable v : i) {
sum = sum + v.get();
}
V.set(sum);
context.write(key, V);
// 執(zhí)行結(jié)果
// i100:i181 1
// i100:i184 2
}
}
}
第三步，對(duì)歌曲id組合列表進(jìn)行計(jì)數(shù)餐抢，建立歌曲id的同現(xiàn)矩陣现使。
運(yùn)行加過如下：
第三步
第四步：
public class Step4 {
public static boolean run(Configuration config, Map<String, String> paths) {
try {
FileSystem fs = FileSystem.get(config);
Job job = Job.getInstance(config);
job.setJobName("step4");
job.setJarByClass(StartRun.class);
job.setMapperClass(Step4_Mapper.class);
job.setReducerClass(Step4_Reducer.class);
job.setMapOutputKeyClass(Text.class);
job.setMapOutputValueClass(Text.class);
FileInputFormat.setInputPaths(job,
new Path[] { new Path(paths.get("Step4Input1")),
new Path(paths.get("Step4Input2")) });
Path outpath = new Path(paths.get("Step4Output"));
if (fs.exists(outpath)) {
fs.delete(outpath, true);
}
FileOutputFormat.setOutputPath(job, outpath);
boolean f = job.waitForCompletion(true);
return f;
} catch (Exception e) {
e.printStackTrace();
}
return false;
}
static class Step4_Mapper extends Mapper<LongWritable, Text, Text, Text> {
private String flag;// A同現(xiàn)矩陣 or B得分矩陣
// 每個(gè)maptask低匙，初始化時(shí)調(diào)用一次
@Override
protected void setup(Context context) throws IOException,
InterruptedException {
FileSplit split = (FileSplit) context.getInputSplit();
flag = split.getPath().getParent().getName();// 判斷讀的數(shù)據(jù)集
System.out.println(flag + "**********************");
}
@Override
protected void map(LongWritable key, Text value, Context context)
throws IOException, InterruptedException {
String[] tokens = Pattern.compile("[\t,]").split(value.toString());
if (flag.equals("step3")) {// 同現(xiàn)矩陣
// 樣本: i100:i181 1
// i100:i184 2
String[] v1 = tokens[0].split(":");
String itemID1 = v1[0];
String itemID2 = v1[1];
String num = tokens[1];
Text k = new Text(itemID1);// 以前一個(gè)歌曲id為key 比如i100
Text v = new Text("A:" + itemID2 + "," + num);// A:i109,1
// 樣本: i100 A:i181,1
context.write(k, v);
} else if (flag.equals("step2")) {// 用戶對(duì)歌曲id喜愛得分矩陣
// 樣本: u24 i64:1,i218:1,i185:1,
String userID = tokens[0];
for (int i = 1; i < tokens.length; i++) {
String[] vector = tokens[i].split(":");
String itemID = vector[0];// 歌曲idid
String pref = vector[1];// 喜愛分?jǐn)?shù)
Text k = new Text(itemID); // 以歌曲id為key 比如：i100
Text v = new Text("B:" + userID + "," + pref); // B:u401,2
// 樣本: i64 B:u24,1
context.write(k, v);
}
}
}
}
static class Step4_Reducer extends Reducer<Text, Text, Text, Text> {
@Override
protected void reduce(Text key, Iterable<Text> values, Context context)
throws IOException, InterruptedException {
// A同現(xiàn)矩陣 or B得分矩陣
// 某一個(gè)歌曲id，針對(duì)它和其他所有歌曲id的同現(xiàn)次數(shù)碳锈，都在mapA集合中
Map<String, Integer> mapA = new HashMap<String, Integer>();
//和該歌曲id（key中的itemID）同現(xiàn)的其他歌曲id的同現(xiàn)集合//
//其他歌曲idID為map的key顽冶，同現(xiàn)數(shù)字為值
Map<String, Integer> mapB = new HashMap<String, Integer>();
//該歌曲id（key中的itemID），所有用戶的推薦權(quán)重分?jǐn)?shù)
for (Text line : values) {
String val = line.toString();
if (val.startsWith("A:")) {// 表示歌曲id同現(xiàn)數(shù)字// 樣本: i100 A:i181,1
String[] kv = Pattern.compile("[\t,]").split(val.substring(2));
try {
mapA.put(kv[0], Integer.parseInt(kv[1]));//mapA:"i181" -> "1"
} catch (Exception e) {
e.printStackTrace();
}
} else if (val.startsWith("B:")) {// 樣本: i64 B:u24,1
String[] kv = Pattern.compile("[\t,]").split(
val.substring(2));
try {
mapB.put(kv[0], Integer.parseInt(kv[1]));//mapB:"u24" -> "1"
} catch (Exception e) {
e.printStackTrace();
}
}
}
double result = 0;
//同現(xiàn)矩陣A
Iterator<String> iter = mapA.keySet().iterator();
//MR原語特征售碳，這里只有一種歌曲的同現(xiàn)列表
while (iter.hasNext()) {
String mapk = iter.next();// itemID
int num = mapA.get(mapk).intValue();
Iterator<String> iterb = mapB.keySet().iterator();
//MR原語特征强重，這里是所有用戶的同一歌曲的評(píng)分，迭代之
while (iterb.hasNext()) {//迭代用戶名
String mapkb = iterb.next();// userID
int pref = mapB.get(mapkb).intValue();
//注意這里的計(jì)算思維理解：
//針對(duì)A歌曲
//使用用戶對(duì)A歌曲的分值
//逐一乘以與A歌曲有同現(xiàn)的歌曲的次數(shù)
//但是計(jì)算推薦向量的時(shí)候需要的是A歌曲同現(xiàn)的歌曲贸人，用同現(xiàn)次數(shù)乘以各自的分值
result = num * pref;// 矩陣乘法相乘計(jì)算
//Text k = new Text(mapkb);
//Text v = new Text(mapk + "," + result);
//結(jié)果樣本: u2723 i9,8.0
//context.write(k, v);
Text k = new Text(mapkb+","+mapk);
Text v = new Text( key.toString() + "," + result);
//key:101
// 結(jié)果樣本: u3,101 101,4.0 *
// 結(jié)果樣本: u3,102 101,4.0
// 結(jié)果樣本: u3,103 101,4.0
//key:102
// 結(jié)果樣本: u3,101 102,4.0 *
// 結(jié)果樣本: u3,102 102,4.0
// 結(jié)果樣本: u3,103 102,4.0
context.write(k, v);
}
}
}
}
}
第四步比較飄
- 把同現(xiàn)矩陣和得分矩陣相乘
- 利用MR原語特征间景，按歌曲分組
- 這樣相同歌曲的同現(xiàn)列表和所有用戶對(duì)該歌曲的評(píng)分進(jìn)到一個(gè)reduce中
這一步也可以理解為，之前邏輯我們是打橫計(jì)算的艺智，現(xiàn)在我們先打豎把所有參數(shù)計(jì)算好拱燃，把整個(gè)矩陣填好，第五步再打橫的一條數(shù)據(jù)一條數(shù)據(jù)技術(shù)出來：
先把整個(gè)矩陣的數(shù)計(jì)算出來
運(yùn)算結(jié)果：
第四步結(jié)果
第五步：
public class Step5 {
private final static Text K = new Text();
private final static Text V = new Text();
public static boolean run(Configuration config, Map<String, String> paths) {
try {
FileSystem fs = FileSystem.get(config);
Job job = Job.getInstance(config);
job.setJobName("step5");
job.setJarByClass(StartRun.class);
job.setMapperClass(Step5_Mapper.class);
job.setReducerClass(Step5_Reducer.class);
job.setMapOutputKeyClass(Text.class);
job.setMapOutputValueClass(Text.class);
FileInputFormat
.addInputPath(job, new Path(paths.get("Step5Input")));
Path outpath = new Path(paths.get("Step5Output"));
if (fs.exists(outpath)) {
fs.delete(outpath, true);
}
FileOutputFormat.setOutputPath(job, outpath);
boolean f = job.waitForCompletion(true);
return f;
} catch (Exception e) {
e.printStackTrace();
}
return false;
}
static class Step5_Mapper extends Mapper<LongWritable, Text, Text, Text> {
/**
* 原封不動(dòng)輸出
*/
@Override
protected void map(LongWritable key, Text value, Context context)
throws IOException, InterruptedException {
// 樣本: u2723:101 i9,8.0
String[] tokens = Pattern.compile("[\t]").split(value.toString());
Text k = new Text(tokens[0]);// 用戶為key
Text v = new Text(tokens[1] );
context.write(k, v);
}
}
static class Step5_Reducer extends Reducer<Text, Text, Text, Text> {
@Override
protected void reduce(Text key, Iterable<Text> values, Context context)
throws IOException, InterruptedException {
Map<String, Double> map = new HashMap<String, Double>();// 結(jié)果
Double score = 0.0;
for (Text line : values) {// i9,4.0
String[] tokens = line.toString().split(",");
// String itemID = tokens[0];
score += Double.parseDouble(tokens[1]);
}
String[] tmp = StringUtils.split(key.toString(),',');
key.set(tmp[0]);
Text v = new Text(tmp[1]+","+String.valueOf(score));
context.write(key, v);
}
// 樣本: u13 i9,5.0
}
}
- 把相乘之后的矩陣相加獲得結(jié)果矩陣
-
還是按用戶分組力惯，將該用戶所有歌曲的推薦向量求和
運(yùn)行結(jié)果：
第五步結(jié)果
第六步：
public class Step6 {
private final static Text K = new Text();
private final static Text V = new Text();
public static boolean run(Configuration config, Map<String, String> paths) {
try {
FileSystem fs = FileSystem.get(config);
Job job = Job.getInstance(config);
job.setJobName("step6");
job.setJarByClass(StartRun.class);
job.setMapperClass(Step6_Mapper.class);
job.setReducerClass(Step6_Reducer.class);
job.setSortComparatorClass(NumSort.class);
job.setGroupingComparatorClass(UserGroup.class);
job.setMapOutputKeyClass(PairWritable.class);
job.setMapOutputValueClass(Text.class);
FileInputFormat
.addInputPath(job, new Path(paths.get("Step6Input")));
Path outpath = new Path(paths.get("Step6Output"));
if (fs.exists(outpath)) {
fs.delete(outpath, true);
}
FileOutputFormat.setOutputPath(job, outpath);
boolean f = job.waitForCompletion(true);
return f;
} catch (Exception e) {
e.printStackTrace();
}
return false;
}
static class Step6_Mapper extends
Mapper<LongWritable, Text, PairWritable, Text> {
@Override
protected void map(LongWritable key, Text value, Context context)
throws IOException, InterruptedException {
String[] tokens = Pattern.compile("[\t,]").split(value.toString());
String u = tokens[0];
String item = tokens[1];
String num = tokens[2];
PairWritable k = new PairWritable();
k.setUid(u);
k.setNum(Double.parseDouble(num));
V.set(item + ":" + num);
context.write(k, V);
}
}
static class Step6_Reducer extends Reducer<PairWritable, Text, Text, Text> {
@Override
protected void reduce(PairWritable key, Iterable<Text> values,
Context context) throws IOException, InterruptedException {
int i = 0;
StringBuffer sb = new StringBuffer();
for (Text v : values) {
if (i == 10)
break;
sb.append(v.toString() + ",");
i++;
}
K.set(key.getUid());
V.set(sb.toString());
context.write(K, V);
}
}
static class PairWritable implements WritableComparable<PairWritable> {
private String uid;
private double num;
@Override
public void write(DataOutput out) throws IOException {
out.writeUTF(uid);
out.writeDouble(num);
}
@Override
public void readFields(DataInput in) throws IOException {
this.uid = in.readUTF();
this.num = in.readDouble();
}
@Override
public int compareTo(PairWritable o) {
int r = this.uid.compareTo(o.getUid());
if (r == 0) {
return Double.compare(this.num, o.getNum());
}
return r;
}
public String getUid() {
return uid;
}
public void setUid(String uid) {
this.uid = uid;
}
public double getNum() {
return num;
}
public void setNum(double num) {
this.num = num;
}
}
static class NumSort extends WritableComparator {
public NumSort() {
super(PairWritable.class, true);
}
@Override
public int compare(WritableComparable a, WritableComparable b) {
PairWritable o1 = (PairWritable) a;
PairWritable o2 = (PairWritable) b;
int r = o1.getUid().compareTo(o2.getUid());
if (r == 0) {
return -Double.compare(o1.getNum(), o2.getNum());
}
return r;
}
}
static class UserGroup extends WritableComparator {
public UserGroup() {
super(PairWritable.class, true);
}
@Override
public int compare(WritableComparable a, WritableComparable b) {
PairWritable o1 = (PairWritable) a;
PairWritable o2 = (PairWritable) b;
return o1.getUid().compareTo(o2.getUid());
}
}
}
第六步碗誉，按照推薦得分降序排序，每個(gè)用戶列出10個(gè)推薦歌曲父晶。
運(yùn)行結(jié)果：
最終結(jié)果
最終結(jié)果哮缺，key作為用戶id，value為推薦分?jǐn)?shù)由高到低的歌曲ID甲喝。
后語：
1.第四步可能有點(diǎn)繞尝苇，看代碼可能看不懂，如果是這樣子的話我建議你拿支筆埠胖，在紙上寫寫那張表的計(jì)算過程糠溜，找找其中規(guī)律，就發(fā)現(xiàn)直撤，妙胺歉汀！
2.用戶的action操作評(píng)分是我暫時(shí)亂寫的谋竖，具體要看你公司業(yè)務(wù)需要红柱，設(shè)計(jì)具體操作的分值。
2.生產(chǎn)中肯定不會(huì)這么簡(jiǎn)單蓖乘，例如要考慮一開始用戶沒數(shù)據(jù)時(shí)個(gè)推薦什么歌曲锤悄；推薦歌曲里不應(yīng)該有已收藏的歌曲，但收藏這個(gè)操作的分值不能低嘉抒，所以要去到下一步的過濾零聚。。。具體怎么優(yōu)化歡迎大家留言~
代碼下載地址：https://github.com/MichaelYipInGitHub/RecommendDemo
