Spark SQL（七）之基于用户的相似度公式

我就是我 2023-01-18 09:27 23阅读 0赞

### 一、基于用户的Jaccard相似度公式 ###

![20210505214149604.png][]

其中，u、v表示任意两个用户，N(u)表示用户u喜欢的物品集合,N(v)表示用户v喜欢物品的集合。

代码

public class UserCFApp {
    
        public static void main(String[]args){
    
            SparkConf sparkConf = new SparkConf();
            sparkConf.setAppName("UserCFApp");
            sparkConf.setMaster("local[*]");
            SparkSession sparkSession = SparkSession.builder().config(sparkConf).getOrCreate();
            
            String url = "jdbc:mysql://localhost:3306/spark-mysql?useUnicode=true&characterEncoding=utf8&autoReconnect=true&failOverReadOnly=false";
            String driver = "com.mysql.jdbc.Driver";
            String user = "root";
            String password = "admin";
            Dataset<Row> dataset = sparkSession.read()
                    .format("jdbc")
                    .option("driver", driver)
                    .option("url",url)
                    .option("dbtable","user_item")
                    .option("user",user)
                    .option("password",password)
                    .load();
            Dataset<Row> userCount = dataset.groupBy("user_id").count();
            Dataset<Row> user2UserCount = dataset.as("a").join(dataset.as("b"),
                    functions.column("a.item_id").$eq$eq$eq(functions.column("b.item_id")))
                    .where(functions.column("a.user_id").notEqual(functions.column("b.user_id")))
                    .select(functions.column("a.user_id").as("a_user_id"),
                            functions.column("b.user_id").as("b_user_id"))
                    .groupBy("a_user_id", "b_user_id").count();
    
            Dataset<Row> result =  user2UserCount.as("u2u")
                     .join(userCount.as("uc1"), functions.column("u2u.a_user_id").$eq$eq$eq(functions.column("uc1.user_id")))
                     .join(userCount.as("uc2"), functions.column("u2u.b_user_id").$eq$eq$eq(functions.column("uc2.user_id")))
             .selectExpr("u2u.a_user_id", "u2u.b_user_id", "u2u.count/(uc1.count + uc2.count - u2u.count) as count");
    
    
            result.show();
    //        result.write()
    //                .mode(SaveMode.Overwrite)
    //                .format("jdbc")
    //                .option("driver", driver)
    //                .option("url",url)
    //                .option("dbtable","user_similar")
    //                .option("user",user)
    //                .option("password",password)
    //                .save();
    
             sparkSession.stop();
        }
    }

### 二、基于用户的余弦相似度公式 ###

![20210505214212626.png][]

其中，u、v表示任意两个用户，N(u)表示用户u喜欢的物品集合,N(v)表示用户v喜欢物品的集合。

public class UserCF2App {
    
        public static void main(String[]args){
    
            SparkConf sparkConf = new SparkConf();
            sparkConf.setAppName("UserCFApp");
            sparkConf.setMaster("local[*]");
            SparkSession sparkSession = SparkSession.builder().config(sparkConf).getOrCreate();
            
            String url = "jdbc:mysql://localhost:3306/spark-mysql?useUnicode=true&characterEncoding=utf8&autoReconnect=true&failOverReadOnly=false";
            String driver = "com.mysql.jdbc.Driver";
            String user = "root";
            String password = "admin";
            Dataset<Row> dataset = sparkSession.read()
                    .format("jdbc")
                    .option("driver", driver)
                    .option("url",url)
                    .option("dbtable","user_item")
                    .option("user",user)
                    .option("password",password)
                    .load();
            Dataset<Row> userCount = dataset.groupBy("user_id").count();
            Dataset<Row> user2UserCount = dataset.as("a").join(dataset.as("b"),
                    functions.column("a.item_id").$eq$eq$eq(functions.column("b.item_id")))
                    .where(functions.column("a.user_id").notEqual(functions.column("b.user_id")))
                    .select(functions.column("a.user_id").as("a_user_id"),
                            functions.column("b.user_id").as("b_user_id"))
                    .groupBy("a_user_id", "b_user_id").count();
    
            Dataset<Row> result =  user2UserCount.as("u2u")
                     .join(userCount.as("uc1"), functions.column("u2u.a_user_id").$eq$eq$eq(functions.column("uc1.user_id")))
                     .join(userCount.as("uc2"), functions.column("u2u.b_user_id").$eq$eq$eq(functions.column("uc2.user_id")))
             .selectExpr("u2u.a_user_id", "u2u.b_user_id", "u2u.count/pow(uc1.count * uc2.count, 0.5) as count");
    
    
            result.show();
    //        result.write()
    //                .mode(SaveMode.Overwrite)
    //                .format("jdbc")
    //                .option("driver", driver)
    //                .option("url",url)
    //                .option("dbtable","user_similar")
    //                .option("user",user)
    //                .option("password",password)
    //                .save();
    
             sparkSession.stop();
        }
    }

[20210505214149604.png]: /images/20221021/353f40fb365d45d197812b0a2046ad52.png
[20210505214212626.png]: /images/20221021/241ed4c0736c4c5c84bfbcd6e27149c1.png