Skip to content

Instantly share code, notes, and snippets.

@aseemk

aseemk/neo4j-cypher-weighted-followers.md

Last active March 21, 2022 13:54
Show Gist options
  • Save aseemk/8049714 to your computer and use it in GitHub Desktop.
Save aseemk/8049714 to your computer and use it in GitHub Desktop.
Download ZIP

Revisions

  1. aseemk revised this gist Feb 9, 2014. 1 changed file with 1 addition and 1 deletion.
    2 changes: 1 addition & 1 deletion neo4j-cypher-weighted-followers.md
    View file
    Original file line number Diff line number Diff line change
    @@ -10,7 +10,7 @@ RETURN user, numFollowers,
    ROUND(totalWeighted * 100 / numFollowers) / 100.0 AS avgFollowerWeight
    ```

    The idea is to weigh followers more who follow fewer *other* people. So for each follower, their "weighted" contribution is `1 / numFollowing`. To prevent division by zero, the original user is included in the "following". The result is that these "normalized" follower counts are always less than users' "simple" follower counts -- but they still make sense relative to each other. Nice!
    The idea is to give more weight to followers who follow fewer *other* people. So for each follower, their "weighted" contribution is `1 / numFollowing`. To prevent division by zero, the original user is included in the "following". The result is that these "normalized" follower counts are always less than users' "simple" follower counts -- but they still make sense relative to each other. Nice!

    One missing feature of the current query is that it ignores users w/ no followers. We can include them by changing the first relationship to an optional match, too, but then we get back to divison by zero. We could always add 1 to numFollowing, but then that gives users with no followers a weighted follower count of 1, and users with actual followers a potentially lower weighted follower count. I haven't figured out how to solve this yet.

  2. aseemk revised this gist Dec 20, 2013. 1 changed file with 1 addition and 1 deletion.
    2 changes: 1 addition & 1 deletion neo4j-cypher-weighted-followers.md
    View file
    Original file line number Diff line number Diff line change
    @@ -10,7 +10,7 @@ RETURN user, numFollowers,
    ROUND(totalWeighted * 100 / numFollowers) / 100.0 AS avgFollowerWeight
    ```

    The idea is to weigh followers more who follow fewer *other* people. So for each follower, their "weighted" contribution is `1 / numFollowing`. To prevent division by zero, the user is included in the "following". The result is that these "normalized" follower counts are always less than users' "simple" follower counts -- but they still make sense relative to each other. Nice!
    The idea is to weigh followers more who follow fewer *other* people. So for each follower, their "weighted" contribution is `1 / numFollowing`. To prevent division by zero, the original user is included in the "following". The result is that these "normalized" follower counts are always less than users' "simple" follower counts -- but they still make sense relative to each other. Nice!

    One missing feature of the current query is that it ignores users w/ no followers. We can include them by changing the first relationship to an optional match, too, but then we get back to divison by zero. We could always add 1 to numFollowing, but then that gives users with no followers a weighted follower count of 1, and users with actual followers a potentially lower weighted follower count. I haven't figured out how to solve this yet.

  3. aseemk revised this gist Dec 20, 2013. 1 changed file with 40 additions and 50 deletions.
    90 changes: 40 additions & 50 deletions neo4j-cypher-weighted-followers.md
    View file
    Original file line number Diff line number Diff line change
    @@ -3,63 +3,53 @@ This is a Neo4j 1.9 (pre-2.0) query:
    ```
    START user=node:nodes(type='user')
    MATCH user <-[:follows]- follower -[?:follows]-> other
    WITH user, follower, 1.0 / (0.0 + COUNT(other)) AS weighted
    WITH user, follower, 1.0 / COUNT(other) AS weighted
    WITH user, COUNT(follower) AS numFollowers, SUM(weighted) as totalWeighted
    RETURN user, numFollowers, ROUND(totalWeighted * 100) / 100.0 AS totalWeighted
    ORDER BY ID(user)
    RETURN user, numFollowers,
    ROUND(totalWeighted * 100) / 100.0 AS totalWeighted,
    ROUND(totalWeighted * 100 / numFollowers) / 100.0 AS avgFollowerWeight
    ```

    The idea is to weigh followers more who follow fewer *other* people. So for each follower, their "weighted" contribution is `1 / numFollowing`. To prevent division by zero, the user is included in the "following". The result is that these "normalized" follower counts are always less than users' "simple" follower counts -- but they still make sense relative to each other. Nice!

    One missing feature of the current query is that it ignores users w/ no followers. We can include them by changing the first relationship to an optional match, too, but then we get back to divison by zero. We could always add 1 to numFollowing, but then that gives users with no followers a weighted follower count of 1, and users with actual followers a potentially lower weighted follower count. I haven't figured out how to solve this yet.

    Run against the current http://node-neo4j-template.herokuapp.com/ database:
    Run against [The Thingdom](http://www.thethingdom.com/)'s database...

    Top 10 users by their "simple" follower count:

    ```
    ==> +----------------------------------------------------------------------------------------------+
    ==> | ID(user) | user.firstName | user.lastName | numFollowers | totalWeighted | avgFollowerWeight |
    ==> +----------------------------------------------------------------------------------------------+
    ==> | 2 | "Aseem" | "Kishore" | 139 | 54.7 | 0.39 |
    ==> | 1 | "Daniel" | "Gasienica" | 72 | 19.06 | 0.26 |
    ==> | 39 | "Jenny" | "Liu" | 40 | 7.46 | 0.19 |
    ==> | 197 | "Ian" | "Gilman" | 22 | 3.41 | 0.15 |
    ==> | 4648 | "The Thingdom" | "" | 20 | 3.32 | 0.17 |
    ==> | 317 | "James" | "Darpinian" | 19 | 2.1 | 0.11 |
    ==> | 443 | "Ben" | "Vanik" | 19 | 2.1 | 0.11 |
    ==> | 125 | "Shelley" | "Gu" | 19 | 2.4 | 0.13 |
    ==> | 6072 | "Jeremy" | "Smith" | 18 | 6.94 | 0.39 |
    ==> | 12186 | "Brad" | "Feld" | 18 | 5.79 | 0.32 |
    ==> +----------------------------------------------------------------------------------------------+
    ```

    Top 10 users by their "weighted" follower count:

    ```
    ==> +-----------------------------------------------------------------+
    ==> | user | numFollowers | totalWeighted |
    ==> +-----------------------------------------------------------------+
    ==> | Node[3]{name:"ii"} | 4 | 1.67 |
    ==> | Node[8]{name:"百度"} | 6 | 2.62 |
    ==> | Node[13]{name:"testme"} | 2 | 0.83 |
    ==> | Node[15]{name:"Jgl"} | 12 | 5.73 |
    ==> | Node[16]{name:"xad2"} | 16 | 6.23 |
    ==> | Node[17]{name:"akbar"} | 6 | 3.67 |
    ==> | Node[19]{name:"lalita"} | 1 | 0.33 |
    ==> | Node[20]{name:"nice-app-thanks"} | 1 | 0.25 |
    ==> | Node[21]{name:"wowsy"} | 8 | 4.31 |
    ==> | Node[23]{name:"jhgfjhgf"} | 10 | 5.87 |
    ==> | Node[24]{name:"suroor"} | 1 | 0.33 |
    ==> | Node[28]{name:"john2"} | 1 | 1.0 |
    ==> | Node[31]{name:"pruebajuandd18"} | 2 | 0.67 |
    ==> | Node[32]{name:"manish"} | 2 | 0.83 |
    ==> | Node[33]{name:"dibz"} | 2 | 0.5 |
    ==> | Node[34]{name:"demian"} | 1 | 0.5 |
    ==> | Node[35]{name:"kumar"} | 6 | 2.84 |
    ==> | Node[36]{name:"oliverjash"} | 1 | 0.25 |
    ==> | Node[38]{name:"Mom"} | 1 | 0.33 |
    ==> | Node[43]{name:" pete"} | 1 | 0.5 |
    ==> | Node[45]{name:"sarah"} | 2 | 0.39 |
    ==> | Node[46]{name:"diman1"} | 1 | 0.33 |
    ==> | Node[48]{name:"tilli"} | 9 | 3.78 |
    ==> | Node[52]{name:"HLASJQ"} | 2 | 1.5 |
    ==> | Node[53]{name:"Momo"} | 8 | 4.0 |
    ==> | Node[54]{name:"pepperone233"} | 2 | 0.45 |
    ==> | Node[57]{name:"ElSebita"} | 7 | 3.68 |
    ==> | Node[58]{name:"rahul"} | 5 | 1.28 |
    ==> | Node[59]{name:"blup"} | 1 | 1.0 |
    ==> | Node[60]{name:"demian2"} | 13 | 7.14 |
    ==> | Node[62]{name:"Purnendu"} | 3 | 1.08 |
    ==> | Node[63]{name:"test"} | 8 | 4.33 |
    ==> | Node[67]{name:"woowee"} | 4 | 2.39 |
    ==> | Node[68]{name:"jimz"} | 1 | 1.0 |
    ==> | Node[73]{name:"ryan"} | 1 | 0.25 |
    ==> | Node[75]{name:"träte2"} | 2 | 0.58 |
    ==> | Node[76]{name:"ohMy"} | 1 | 0.5 |
    ==> | Node[82]{name:"alex"} | 1 | 0.25 |
    ==> | Node[85]{name:"fremNOKdegroBO"} | 4 | 1.08 |
    ==> | Node[90]{name:"teest"} | 6 | 2.03 |
    ==> | Node[93]{name:"tasinet"} | 1 | 0.5 |
    ==> | Node[94]{name:"Щы"} | 1 | 0.17 |
    ==> +-----------------------------------------------------------------+
    ==> +----------------------------------------------------------------------------------------------+
    ==> | ID(user) | user.firstName | user.lastName | numFollowers | totalWeighted | avgFollowerWeight |
    ==> +----------------------------------------------------------------------------------------------+
    ==> | 2 | "Aseem" | "Kishore" | 139 | 54.7 | 0.39 |
    ==> | 1 | "Daniel" | "Gasienica" | 72 | 19.06 | 0.26 |
    ==> | 39 | "Jenny" | "Liu" | 40 | 7.46 | 0.19 |
    ==> | 6072 | "Jeremy" | "Smith" | 18 | 6.94 | 0.39 |
    ==> | 12186 | "Brad" | "Feld" | 18 | 5.79 | 0.32 |
    ==> | 197 | "Ian" | "Gilman" | 22 | 3.41 | 0.15 |
    ==> | 4648 | "The Thingdom" | "" | 20 | 3.32 | 0.17 |
    ==> | 200 | "Frida" | "Kumar" | 8 | 3.23 | 0.4 |
    ==> | 3451 | "Anh Huy" | "Truong" | 7 | 2.68 | 0.38 |
    ==> | 11855 | "Andrew" | "Dorne" | 8 | 2.46 | 0.31 |
    ==> +----------------------------------------------------------------------------------------------+
    ```
  4. aseemk revised this gist Dec 20, 2013. 1 changed file with 1 addition and 2 deletions.
    3 changes: 1 addition & 2 deletions neo4j-cypher-weighted-followers.md
    View file
    Original file line number Diff line number Diff line change
    @@ -4,8 +4,7 @@ This is a Neo4j 1.9 (pre-2.0) query:
    START user=node:nodes(type='user')
    MATCH user <-[:follows]- follower -[?:follows]-> other
    WITH user, follower, 1.0 / (0.0 + COUNT(other)) AS weighted
    WITH user, COUNT(follower) AS numFollowers,
    REDUCE(total = 0, w IN COLLECT(weighted) : total + w) AS totalWeighted
    WITH user, COUNT(follower) AS numFollowers, SUM(weighted) as totalWeighted
    RETURN user, numFollowers, ROUND(totalWeighted * 100) / 100.0 AS totalWeighted
    ORDER BY ID(user)
    ```
  5. aseemk renamed this gist Dec 20, 2013. 1 changed file with 0 additions and 0 deletions.
    0 example.md → neo4j-cypher-weighted-followers.md
    View file
    File renamed without changes.
  6. aseemk revised this gist Dec 20, 2013. 1 changed file with 2 additions and 12 deletions.
    14 changes: 2 additions & 12 deletions example.md
    View file
    Original file line number Diff line number Diff line change
    @@ -10,19 +10,9 @@ RETURN user, numFollowers, ROUND(totalWeighted * 100) / 100.0 AS totalWeighted
    ORDER BY ID(user)
    ```

    The idea is to weigh followers more who follow fewer *other* people.
    So for each follower, their "weighted" contribution is `1 / numFollowing`.
    To prevent division by zero, the user is included in the "following".
    The idea is to weigh followers more who follow fewer *other* people. So for each follower, their "weighted" contribution is `1 / numFollowing`. To prevent division by zero, the user is included in the "following". The result is that these "normalized" follower counts are always less than users' "simple" follower counts -- but they still make sense relative to each other. Nice!

    The result is that this "normalized" follower count is always less than your
    "simple" follower count -- but roughly in the same ballpark! Nice.

    One missing feature of the current query is that it ignores users w/ no followers.
    We can include them by changing the first relationship to an optional match, too,
    but then we get back to divison by zero. We could always add 1 to numFollowing,
    but then that gives users with no followers a weighted follower count of 1,
    and users with actual followers a potentially lower weighted follower count.
    Haven't figured out how to solve this yet.
    One missing feature of the current query is that it ignores users w/ no followers. We can include them by changing the first relationship to an optional match, too, but then we get back to divison by zero. We could always add 1 to numFollowing, but then that gives users with no followers a weighted follower count of 1, and users with actual followers a potentially lower weighted follower count. I haven't figured out how to solve this yet.

    Run against the current http://node-neo4j-template.herokuapp.com/ database:

  7. aseemk renamed this gist Dec 20, 2013. 1 changed file with 5 additions and 1 deletion.
    6 changes: 5 additions & 1 deletion cypher.txt → example.md
    View file
    Original file line number Diff line number Diff line change
    @@ -1,15 +1,17 @@
    This is a Neo4j 1.9 (pre-2.0) query:

    ```
    START user=node:nodes(type='user')
    MATCH user <-[:follows]- follower -[?:follows]-> other
    WITH user, follower, 1.0 / (0.0 + COUNT(other)) AS weighted
    WITH user, COUNT(follower) AS numFollowers,
    REDUCE(total = 0, w IN COLLECT(weighted) : total + w) AS totalWeighted
    RETURN user, numFollowers, ROUND(totalWeighted * 100) / 100.0 AS totalWeighted
    ORDER BY ID(user)
    ```

    The idea is to weigh followers more who follow fewer *other* people.
    So for each follower, their "weighted" contribution is 1 / numFollowing.
    So for each follower, their "weighted" contribution is `1 / numFollowing`.
    To prevent division by zero, the user is included in the "following".

    The result is that this "normalized" follower count is always less than your
    @@ -24,6 +26,7 @@ Haven't figured out how to solve this yet.

    Run against the current http://node-neo4j-template.herokuapp.com/ database:

    ```
    ==> +-----------------------------------------------------------------+
    ==> | user | numFollowers | totalWeighted |
    ==> +-----------------------------------------------------------------+
    @@ -70,3 +73,4 @@ Run against the current http://node-neo4j-template.herokuapp.com/ database:
    ==> | Node[93]{name:"tasinet"} | 1 | 0.5 |
    ==> | Node[94]{name:"Щы"} | 1 | 0.17 |
    ==> +-----------------------------------------------------------------+
    ```
  8. aseemk revised this gist Dec 20, 2013. 1 changed file with 1 addition and 1 deletion.
    2 changes: 1 addition & 1 deletion cypher.txt
    View file
    Original file line number Diff line number Diff line change
    @@ -28,7 +28,7 @@ Run against the current http://node-neo4j-template.herokuapp.com/ database:
    ==> | user | numFollowers | totalWeighted |
    ==> +-----------------------------------------------------------------+
    ==> | Node[3]{name:"ii"} | 4 | 1.67 |
    ==> | Node[8]{name:"百度"} | 6 | 2.62 |
    ==> | Node[8]{name:"百度"} | 6 | 2.62 |
    ==> | Node[13]{name:"testme"} | 2 | 0.83 |
    ==> | Node[15]{name:"Jgl"} | 12 | 5.73 |
    ==> | Node[16]{name:"xad2"} | 16 | 6.23 |
  9. aseemk revised this gist Dec 20, 2013. 1 changed file with 1 addition and 1 deletion.
    2 changes: 1 addition & 1 deletion cypher.txt
    View file
    Original file line number Diff line number Diff line change
    @@ -28,7 +28,7 @@ Run against the current http://node-neo4j-template.herokuapp.com/ database:
    ==> | user | numFollowers | totalWeighted |
    ==> +-----------------------------------------------------------------+
    ==> | Node[3]{name:"ii"} | 4 | 1.67 |
    ==> | Node[8]{name:"百度"} | 6 | 2.62 |
    ==> | Node[8]{name:"百度"} | 6 | 2.62 |
    ==> | Node[13]{name:"testme"} | 2 | 0.83 |
    ==> | Node[15]{name:"Jgl"} | 12 | 5.73 |
    ==> | Node[16]{name:"xad2"} | 16 | 6.23 |
  10. aseemk revised this gist Dec 20, 2013. 1 changed file with 49 additions and 49 deletions.
    98 changes: 49 additions & 49 deletions cypher.txt
    View file
    Original file line number Diff line number Diff line change
    @@ -2,10 +2,10 @@ This is a Neo4j 1.9 (pre-2.0) query:

    START user=node:nodes(type='user')
    MATCH user <-[:follows]- follower -[?:follows]-> other
    WITH user, follower, 1.0 / COUNT(other) AS weighted
    WITH user, follower, 1.0 / (0.0 + COUNT(other)) AS weighted
    WITH user, COUNT(follower) AS numFollowers,
    REDUCE(total = 0, w IN COLLECT(weighted) : total + w) AS totalWeighted
    RETURN user, numFollowers, ROUND(totalWeighted), totalWeighted
    REDUCE(total = 0, w IN COLLECT(weighted) : total + w) AS totalWeighted
    RETURN user, numFollowers, ROUND(totalWeighted * 100) / 100.0 AS totalWeighted
    ORDER BY ID(user)

    The idea is to weigh followers more who follow fewer *other* people.
    @@ -24,49 +24,49 @@ Haven't figured out how to solve this yet.

    Run against the current http://node-neo4j-template.herokuapp.com/ database:

    ==> +----------------------------------------------------------------------------------------------+
    ==> | user | numFollowers | ROUND(totalWeighted) | totalWeighted |
    ==> +----------------------------------------------------------------------------------------------+
    ==> | Node[3]{name:"ii"} | 4 | 2 | 1.6666666666666665 |
    ==> | Node[8]{name:"百度"} | 6 | 3 | 2.6166666666666667 |
    ==> | Node[13]{name:"testme"} | 2 | 1 | 0.8333333333333333 |
    ==> | Node[15]{name:"Jgl"} | 12 | 6 | 5.726190476190476 |
    ==> | Node[16]{name:"xad2"} | 16 | 6 | 6.233333333333334 |
    ==> | Node[17]{name:"akbar"} | 6 | 4 | 3.6666666666666665 |
    ==> | Node[19]{name:"lalita"} | 1 | 0 | 0.3333333333333333 |
    ==> | Node[20]{name:"nice-app-thanks"} | 1 | 0 | 0.25 |
    ==> | Node[21]{name:"wowsy"} | 8 | 4 | 4.30952380952381 |
    ==> | Node[23]{name:"jhgfjhgf"} | 10 | 6 | 5.866666666666666 |
    ==> | Node[24]{name:"suroor"} | 1 | 0 | 0.3333333333333333 |
    ==> | Node[28]{name:"john2"} | 1 | 1 | 1.0 |
    ==> | Node[31]{name:"pruebajuandd18"} | 2 | 1 | 0.6666666666666666 |
    ==> | Node[32]{name:"manish"} | 2 | 1 | 0.8333333333333333 |
    ==> | Node[33]{name:"dibz"} | 2 | 1 | 0.5 |
    ==> | Node[34]{name:"demian"} | 1 | 1 | 0.5 |
    ==> | Node[35]{name:"kumar"} | 6 | 3 | 2.842857142857143 |
    ==> | Node[36]{name:"oliverjash"} | 1 | 0 | 0.25 |
    ==> | Node[38]{name:"Mom"} | 1 | 0 | 0.3333333333333333 |
    ==> | Node[43]{name:" pete"} | 1 | 1 | 0.5 |
    ==> | Node[45]{name:"sarah"} | 2 | 0 | 0.39285714285714285 |
    ==> | Node[46]{name:"diman1"} | 1 | 0 | 0.3333333333333333 |
    ==> | Node[48]{name:"tilli"} | 9 | 4 | 3.783333333333333 |
    ==> | Node[52]{name:"HLASJQ"} | 2 | 2 | 1.5 |
    ==> | Node[53]{name:"Momo"} | 8 | 4 | 4.0 |
    ==> | Node[54]{name:"pepperone233"} | 2 | 0 | 0.45 |
    ==> | Node[57]{name:"ElSebita"} | 7 | 4 | 3.6761904761904765 |
    ==> | Node[58]{name:"rahul"} | 5 | 1 | 1.2833333333333334 |
    ==> | Node[59]{name:"blup"} | 1 | 1 | 1.0 |
    ==> | Node[60]{name:"demian2"} | 13 | 7 | 7.142857142857142 |
    ==> | Node[62]{name:"Purnendu"} | 3 | 1 | 1.0833333333333333 |
    ==> | Node[63]{name:"test"} | 8 | 4 | 4.333333333333333 |
    ==> | Node[67]{name:"woowee"} | 4 | 2 | 2.392857142857143 |
    ==> | Node[68]{name:"jimz"} | 1 | 1 | 1.0 |
    ==> | Node[73]{name:"ryan"} | 1 | 0 | 0.25 |
    ==> | Node[75]{name:"träte2"} | 2 | 1 | 0.5833333333333333 |
    ==> | Node[76]{name:"ohMy"} | 1 | 1 | 0.5 |
    ==> | Node[82]{name:"alex"} | 1 | 0 | 0.25 |
    ==> | Node[85]{name:"fremNOKdegroBO"} | 4 | 1 | 1.0833333333333333 |
    ==> | Node[90]{name:"teest"} | 6 | 2 | 2.033333333333333 |
    ==> | Node[93]{name:"tasinet"} | 1 | 1 | 0.5 |
    ==> | Node[94]{name:"Щы"} | 1 | 0 | 0.16666666666666666 |
    ==> +----------------------------------------------------------------------------------------------+
    ==> +-----------------------------------------------------------------+
    ==> | user | numFollowers | totalWeighted |
    ==> +-----------------------------------------------------------------+
    ==> | Node[3]{name:"ii"} | 4 | 1.67 |
    ==> | Node[8]{name:"百度"} | 6 | 2.62 |
    ==> | Node[13]{name:"testme"} | 2 | 0.83 |
    ==> | Node[15]{name:"Jgl"} | 12 | 5.73 |
    ==> | Node[16]{name:"xad2"} | 16 | 6.23 |
    ==> | Node[17]{name:"akbar"} | 6 | 3.67 |
    ==> | Node[19]{name:"lalita"} | 1 | 0.33 |
    ==> | Node[20]{name:"nice-app-thanks"} | 1 | 0.25 |
    ==> | Node[21]{name:"wowsy"} | 8 | 4.31 |
    ==> | Node[23]{name:"jhgfjhgf"} | 10 | 5.87 |
    ==> | Node[24]{name:"suroor"} | 1 | 0.33 |
    ==> | Node[28]{name:"john2"} | 1 | 1.0 |
    ==> | Node[31]{name:"pruebajuandd18"} | 2 | 0.67 |
    ==> | Node[32]{name:"manish"} | 2 | 0.83 |
    ==> | Node[33]{name:"dibz"} | 2 | 0.5 |
    ==> | Node[34]{name:"demian"} | 1 | 0.5 |
    ==> | Node[35]{name:"kumar"} | 6 | 2.84 |
    ==> | Node[36]{name:"oliverjash"} | 1 | 0.25 |
    ==> | Node[38]{name:"Mom"} | 1 | 0.33 |
    ==> | Node[43]{name:" pete"} | 1 | 0.5 |
    ==> | Node[45]{name:"sarah"} | 2 | 0.39 |
    ==> | Node[46]{name:"diman1"} | 1 | 0.33 |
    ==> | Node[48]{name:"tilli"} | 9 | 3.78 |
    ==> | Node[52]{name:"HLASJQ"} | 2 | 1.5 |
    ==> | Node[53]{name:"Momo"} | 8 | 4.0 |
    ==> | Node[54]{name:"pepperone233"} | 2 | 0.45 |
    ==> | Node[57]{name:"ElSebita"} | 7 | 3.68 |
    ==> | Node[58]{name:"rahul"} | 5 | 1.28 |
    ==> | Node[59]{name:"blup"} | 1 | 1.0 |
    ==> | Node[60]{name:"demian2"} | 13 | 7.14 |
    ==> | Node[62]{name:"Purnendu"} | 3 | 1.08 |
    ==> | Node[63]{name:"test"} | 8 | 4.33 |
    ==> | Node[67]{name:"woowee"} | 4 | 2.39 |
    ==> | Node[68]{name:"jimz"} | 1 | 1.0 |
    ==> | Node[73]{name:"ryan"} | 1 | 0.25 |
    ==> | Node[75]{name:"träte2"} | 2 | 0.58 |
    ==> | Node[76]{name:"ohMy"} | 1 | 0.5 |
    ==> | Node[82]{name:"alex"} | 1 | 0.25 |
    ==> | Node[85]{name:"fremNOKdegroBO"} | 4 | 1.08 |
    ==> | Node[90]{name:"teest"} | 6 | 2.03 |
    ==> | Node[93]{name:"tasinet"} | 1 | 0.5 |
    ==> | Node[94]{name:"Щы"} | 1 | 0.17 |
    ==> +-----------------------------------------------------------------+
  11. aseemk revised this gist Dec 20, 2013. 1 changed file with 1 addition and 1 deletion.
    2 changes: 1 addition & 1 deletion cypher.txt
    View file
    Original file line number Diff line number Diff line change
    @@ -28,7 +28,7 @@ Run against the current http://node-neo4j-template.herokuapp.com/ database:
    ==> | user | numFollowers | ROUND(totalWeighted) | totalWeighted |
    ==> +----------------------------------------------------------------------------------------------+
    ==> | Node[3]{name:"ii"} | 4 | 2 | 1.6666666666666665 |
    ==> | Node[8]{name:"百度"} | 6 | 3 | 2.6166666666666667 |
    ==> | Node[8]{name:"百度"} | 6 | 3 | 2.6166666666666667 |
    ==> | Node[13]{name:"testme"} | 2 | 1 | 0.8333333333333333 |
    ==> | Node[15]{name:"Jgl"} | 12 | 6 | 5.726190476190476 |
    ==> | Node[16]{name:"xad2"} | 16 | 6 | 6.233333333333334 |
  12. aseemk created this gist Dec 20, 2013.
    72 changes: 72 additions & 0 deletions cypher.txt
    View file
    Original file line number Diff line number Diff line change
    @@ -0,0 +1,72 @@
    This is a Neo4j 1.9 (pre-2.0) query:

    START user=node:nodes(type='user')
    MATCH user <-[:follows]- follower -[?:follows]-> other
    WITH user, follower, 1.0 / COUNT(other) AS weighted
    WITH user, COUNT(follower) AS numFollowers,
    REDUCE(total = 0, w IN COLLECT(weighted) : total + w) AS totalWeighted
    RETURN user, numFollowers, ROUND(totalWeighted), totalWeighted
    ORDER BY ID(user)

    The idea is to weigh followers more who follow fewer *other* people.
    So for each follower, their "weighted" contribution is 1 / numFollowing.
    To prevent division by zero, the user is included in the "following".

    The result is that this "normalized" follower count is always less than your
    "simple" follower count -- but roughly in the same ballpark! Nice.

    One missing feature of the current query is that it ignores users w/ no followers.
    We can include them by changing the first relationship to an optional match, too,
    but then we get back to divison by zero. We could always add 1 to numFollowing,
    but then that gives users with no followers a weighted follower count of 1,
    and users with actual followers a potentially lower weighted follower count.
    Haven't figured out how to solve this yet.

    Run against the current http://node-neo4j-template.herokuapp.com/ database:

    ==> +----------------------------------------------------------------------------------------------+
    ==> | user | numFollowers | ROUND(totalWeighted) | totalWeighted |
    ==> +----------------------------------------------------------------------------------------------+
    ==> | Node[3]{name:"ii"} | 4 | 2 | 1.6666666666666665 |
    ==> | Node[8]{name:"百度"} | 6 | 3 | 2.6166666666666667 |
    ==> | Node[13]{name:"testme"} | 2 | 1 | 0.8333333333333333 |
    ==> | Node[15]{name:"Jgl"} | 12 | 6 | 5.726190476190476 |
    ==> | Node[16]{name:"xad2"} | 16 | 6 | 6.233333333333334 |
    ==> | Node[17]{name:"akbar"} | 6 | 4 | 3.6666666666666665 |
    ==> | Node[19]{name:"lalita"} | 1 | 0 | 0.3333333333333333 |
    ==> | Node[20]{name:"nice-app-thanks"} | 1 | 0 | 0.25 |
    ==> | Node[21]{name:"wowsy"} | 8 | 4 | 4.30952380952381 |
    ==> | Node[23]{name:"jhgfjhgf"} | 10 | 6 | 5.866666666666666 |
    ==> | Node[24]{name:"suroor"} | 1 | 0 | 0.3333333333333333 |
    ==> | Node[28]{name:"john2"} | 1 | 1 | 1.0 |
    ==> | Node[31]{name:"pruebajuandd18"} | 2 | 1 | 0.6666666666666666 |
    ==> | Node[32]{name:"manish"} | 2 | 1 | 0.8333333333333333 |
    ==> | Node[33]{name:"dibz"} | 2 | 1 | 0.5 |
    ==> | Node[34]{name:"demian"} | 1 | 1 | 0.5 |
    ==> | Node[35]{name:"kumar"} | 6 | 3 | 2.842857142857143 |
    ==> | Node[36]{name:"oliverjash"} | 1 | 0 | 0.25 |
    ==> | Node[38]{name:"Mom"} | 1 | 0 | 0.3333333333333333 |
    ==> | Node[43]{name:" pete"} | 1 | 1 | 0.5 |
    ==> | Node[45]{name:"sarah"} | 2 | 0 | 0.39285714285714285 |
    ==> | Node[46]{name:"diman1"} | 1 | 0 | 0.3333333333333333 |
    ==> | Node[48]{name:"tilli"} | 9 | 4 | 3.783333333333333 |
    ==> | Node[52]{name:"HLASJQ"} | 2 | 2 | 1.5 |
    ==> | Node[53]{name:"Momo"} | 8 | 4 | 4.0 |
    ==> | Node[54]{name:"pepperone233"} | 2 | 0 | 0.45 |
    ==> | Node[57]{name:"ElSebita"} | 7 | 4 | 3.6761904761904765 |
    ==> | Node[58]{name:"rahul"} | 5 | 1 | 1.2833333333333334 |
    ==> | Node[59]{name:"blup"} | 1 | 1 | 1.0 |
    ==> | Node[60]{name:"demian2"} | 13 | 7 | 7.142857142857142 |
    ==> | Node[62]{name:"Purnendu"} | 3 | 1 | 1.0833333333333333 |
    ==> | Node[63]{name:"test"} | 8 | 4 | 4.333333333333333 |
    ==> | Node[67]{name:"woowee"} | 4 | 2 | 2.392857142857143 |
    ==> | Node[68]{name:"jimz"} | 1 | 1 | 1.0 |
    ==> | Node[73]{name:"ryan"} | 1 | 0 | 0.25 |
    ==> | Node[75]{name:"träte2"} | 2 | 1 | 0.5833333333333333 |
    ==> | Node[76]{name:"ohMy"} | 1 | 1 | 0.5 |
    ==> | Node[82]{name:"alex"} | 1 | 0 | 0.25 |
    ==> | Node[85]{name:"fremNOKdegroBO"} | 4 | 1 | 1.0833333333333333 |
    ==> | Node[90]{name:"teest"} | 6 | 2 | 2.033333333333333 |
    ==> | Node[93]{name:"tasinet"} | 1 | 1 | 0.5 |
    ==> | Node[94]{name:"Щы"} | 1 | 0 | 0.16666666666666666 |
    ==> +----------------------------------------------------------------------------------------------+