Human Capital Management Blogs by SAP
Get insider info on SAP SuccessFactors HCM suite for core HR and payroll, time and attendance, talent management, employee experience management, and more in this SAP blog.
Showing results for 
Search instead for 
Did you mean: 
Product and Topic Expert
Product and Topic Expert
注記:当ブログは、SAP アドベントカレンダー2020 の12月24日記事として作成されました。

Note: The blog is written as the article "December 24 of SAP Advent Calendar 2020".

SAP SuccessFactors (以下、SFSF) Learning Management System (以下、LMS) に、Machine Learning (以下ML) 機能 Learning Recommendations がリリースされました。

In 2020, one of the Machine Learning (hereby ML) feature, named Learning Recommendations (hereby LR) is released in Learning Management System (LMS) of SAP SuccessFactors (SFSF).


In this blog, I will introduce the 1) solution overview, 2) set up, 3) feedback.


ソリューションの概要 Solution Overview

SFSF には、複数のレコメンデーション機能があります。

In the SFSF, there are some recommendation features are released or planned.

今回紹介するLR は「パーソナライズド学習コンテンツ提案」と呼ばれています。

The LR is the one of the recommendations called as "Personalized Recommendations".

SFSF LMS に格納されている膨大な学習コンテンツを2つの側面から機械学習のアルゴリズムを使って提案します。(以下、SAPソリューション紹介スライドより抜粋。)

LR recommend the large amount of Learning Items stored in the SFSF LMS using ML algorism with two aspects. (You can see the below pictures from SAP  solution overview slides.)

他のMLソリューションと同じく、トレーニングプロセスとインフェレンスプロセスに分かれています。SFSFはクラウドサービスであるため、On Premise のS/4HANA で実行したCash Application とは異なり、データ抽出やトレーニングについてはSFSFチームで実施します。

As same as the other ML solutions, there are two processes of training the model with employee's learning history and/or profile, topics, etc., and inference to propose the learning contents.

Different from the Cash Application in On Premise S/4HANA environment, the extraction of the training data and training the ML model are done by the SFSF team.


The proposal to be shown on the screen as below, when the model is trained.


You can control which employee can get access to this recommendation can be set by permissions.


(1) 同僚が受けたコースを提案 Recommend the courses which your peer taken

その人の同僚をどのように判別しているのかという点については、SFSF LMSから送られた以下のユーザ属性情報から知ることができます。(更新:2021年1月8日 2018 Q3 release -> 2020 Q3 release )

How the Learning Recommender identified the Peer is the below User Attribute information sent from SFSF LMS. (updated on Jan 8, 2021 from Q3 2018 rel. to Q3 2020 rel.)


You can also find the detail in online help.

(2) 購読するトピックから提案 Recommendations from your subscribed topics


Employee or LMS Admin can add subscription of the topics predefined by SAP.


サポート言語 Support Languages

ラテン言語に加えて、2020 年 11月リリースから日本がサポートされました!

Now Japanese is supported as of November release in 2020 in addition to Latin languages!


設定 Setting

(更新: 2021 年1月8日)

  1. Enabling Recommendations Tile (必須)

  2. Enabling Recommendations Engine (必須)

  3. Enabling Recommendations Newsletter

  4. Enabling Custom Fields of Users to be included in Personalized Recommendations

  5. Assigning Permissions to Users for Accessing Recommendations (必須)

  6. Editing the Legal Disclaimer for Learning Recommendations



  1. Enabling Recommendations Tile (mandatory)

  2. Enabling Recommendations Engine (mandatory)

  3. Enabling Recommendations Newsletter

  4. Enabling Custom Fields of Users to be included in Personalized Recommendations

  5. Assigning Permissions to Users for Accessing Recommendations (mandatory)

  6. Editing the Legal Disclaimer for Learning Recommendations

Please also refer the online help for the setting.


触ってみた感想 Feedback

Learning Content の Topic の抽出について

Learning Recommender は、英語を基本として動作するMLアプリケーションです。









Peer Recommendation について








About extracting Topic of Learning Content

Learning Recommender is an ML application that operates based on English.

Therefore, the topic is extracted after the Japanese content title and description are translated into English by machine translation.

As a result of topic extraction from a certain customer's data, it seems that about 70% is correct, although there were some misjudgments.

First of all, I felt that the title and description of the learning content needed enough detail or words like tags that could identify what kind of content it was, rather than abbreviations.

Please note that customer-specific words (product names, etc.) cannot be identified as topics.

Topics are selected from a list created by SAP. The content of the topic will be updated from time to time and at the request of the customer.

It seems that maintenance of learning content on the customer side will be necessary.

There seems that some improvement is planned in the future, such as not displaying the content of 5 years ago even if there are many students taken. Even so, if a new version of the same content comes out, it is necessary to take measures to invalidate the old content, and put relationship between them or use versioning.

About Peer Recommendation

The peer is identified based on the code in the user profile. Even if the field should contain the code contains text, the LR considers that text as just code and the matching attribute to be a colleague (Peer).

Because the semantic context could not be determined, technical learning items were sometimes proposed to the HR colleagues. It was questioned by the customer but when I checked with the development, there was also a leaning history of the other HR colleague in the past.

In the other case, it is not possible to identify and propose the person in charge of a specific product in the same department.

Please understand that LR can only identify the peer by the user attribute information available in the system. If that specific responsibility is assigned as a custom field to be sent to LR, it might be possible to identify the peer, but in this time the customer does not allocate meaningful custom field, and was not able to check.

If the user is not interested in the recommended learning content, or if it is in his/her learning history in the past, it is not recommended, so I hope that LR gradually to be smarter by using it.