Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Simulated Analysis of MAUT Collaborative Filtering for Learning Object Recommendation, Manouselis+, Social Information Retrieval for Technology-Enhanced Learning & Exchange, 2007 #292

Open
AkihikoWatanabe opened this issue Dec 22, 2018 · 2 comments

Comments

@AkihikoWatanabe
Copy link
Owner

https://pdfs.semanticscholar.org/e5ce/85164f062bdba5e1292bca94bfca0cd2c36f.pdf

@AkihikoWatanabe
Copy link
Owner Author

教員に対して教材を推薦しようという試み(学生ではないようだ)。
教員は、learning resourcesに対して、multi-criteriaなratingを付与することができ、それをCFで活用する(CELEBRATE web portalというヨーロッパのポータルを使用したらしい)。
CFはmemory-basedな手法を使用。target userがあるアイテムを、それぞれのattributeの観点からどのようにratingするかをattributeごとに別々に予測。各attributeのスコアを最終的に統合(元の論文ではただのスコアの足し合わせ)して、推薦スコアとする。

以下が調査された:

  1. ユーザ間の距離の測り方(ユークリッド距離、cossim、ピアソンの相関係数)
  2. neighborsの選び方(定義しておいた最大人数か、相関の重みで選ぶか)
  3. neighborのratingをどのように組み合わせるか(平均、重み付き平均、mean formulaからのdeviation)

�評価する際は、ratingのデータを training/test 80%/20%に分割。テストセットのアイテムに対して、ユーザがratingした情報をどれだけ正しく予測できるかで検証(511 evaluation in test, 2043 evaluations in training)。

ratingのMAE, coverage, アルゴリズムの実行時間で評価。

CorrerationWeightThresholdが各種アルゴリズムで安定した性能。Maximum Number Userはばらつきがでかい。いい感じの設定がみつかれば、Maximum Number Userの方がMAEの観点からは強い。
top-10のアイテムをselectするようにしたら、60%のcoverageになった。
(アルゴリズムの実行時間は、2000程度のevaluationデータに対して、2.5GHZ CPU, 256MEMで20秒とかかかってる。)

@AkihikoWatanabe
Copy link
Owner Author

Learning Resource Exchangeの文脈で使われることを想定(このシステムではヨーロッパのK-12)。

教員による教材のmulti-criteriaのratingは5-scaleで行われた。
どういうcriteriaに対してratingされたかが書いてない。

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

1 participant