Recommendation system of online education

There are two main types of recommendation needs in online education products , subject 、 Video and other educational resources Tutor's recommendation , These two categories of recommendations are to match the resources on the platform with the actual needs . The following topics are mainly discussed 、 Video and other educational resources should be recommended as an example .

actually , Recommendation system is the link between users and platform resource pool , It has a small information load for users , Recommend the most reasonable resources to users , And then to achieve the purpose of improving points and efficiency for users .

In the design of recommendation algorithm , At present, there are mainly recommendations based on users' historical behavior and content .

Recommendations based on users' historical behavior ： There are three main types of user behavior tracks on the platform

1、 Search topic 、 video

2、 Watch Video , download 、 Complete the topic

3、 Previous examinations 、 Record the title of homework , This part is the most detailed behavior record source , Specifically, it will involve all the examinations since users joined the platform 、 Record of the homework done 、 The correctness of the title 、 questions 、 Details of knowledge points and other relevant information .

among , Most of the user behavior is mainly the third kind of user behavior .

The principles of recommendation based on user's historical behavior and content-based recommendation will not be described again , Here we mainly talk about the logic of selecting recommendation strategies ：

According to the specific use scenario , The weights of hybrid recommendations based on users' historical behavior and content are different .

1、 For example, when users search and watch , Due to the strong purpose of users , Have an immediate impact on a certain type of content 、 Strong demand , In making recommendations , It will be mainly based on the user's search clicks 、 Watch the content of the video 、 Topic, etc. for content-based recommendation . meanwhile , Based on the consideration of smoothness and diversity of recommended content , Some recommendation results based on users' historical behavior will be incorporated .

2、 Recommendation based on users' historical behavior is mainly in the scenario of generating personalized counseling programs for users , At this time, the starting point of recommendation is mainly to check the omissions and make up the deficiencies of users in the past period of study 、 Summarize and improve . therefore , Recommendations based on users' historical behavior will better meet users' needs . Considering the timeliness of historical behavior , The recommendation system generates weekly coaching programs as needed , Monthly coaching program 、 Half semester tutoring program and semester tutoring program .

It's worth noting that , Because of the objectivity of exam oriented education , The difficulty of the questions in previous exams 、 Composition of topic content 、 The distribution of different difficulty and question types of the whole set of examination papers has certain stress , therefore , The recommendation system not only needs to generate a certain scale of recommended topics with the user's historical behavior records , We also need to identify the test strategy and test trend from the historical tests , Then, based on the recommendation of users' historical behavior, a set of test questions that are more in line with the actual needs , For user's use . This final recommendation is not only difficult 、 Knowledge points meet the needs of users , In the style of the test paper, it also conforms to the user's usual question making style , At the same time, in the question type 、 The examination strategy is also more in line with the examination style and outline requirements of examination oriented education .

You can see , The recommendation algorithm should not only be able to accurately mine the needs of users , Make personalized recommendations 、 One thousand thousand , At the same time, it should be combined with specific recommended scenarios 、 Grasp the user's usage habits .

Differences with e-commerce recommendation systems ：
Different from the traditional e-commerce recommendation system , The e-commerce recommendation system focuses on improving the conversion rate of users' purchases 、 Mining long tail goods to improve the revenue of enterprises . The education recommendation system needs to pay attention to how to promote users ( Student ) grow up 、 Improve your academic performance , It can be said that this is a very indirect 、 Less explicit optimization objectives . The performance of such an educational recommendation system needs attention in the short term 1、 Download user ratio ( The higher it is, the more people need it )2、 Feedback upload rate ( The higher the rate of feedback uploading , Explain that the recommended content is more appropriate )3、 Feedback evaluation label ： Too hard 、 Too simple 、 Repeat 、 Strange content and other roast options Interact with users , Perceive user needs in the long run , The ultimate goal is to improve students' learning level .
From the perspective of human growth , Learning is good 、 Other abilities are also good , We need to constantly step out of the comfort zone and enter the discomfort zone , It's a painful process .
The difficulty of recommendation also lies in this , We should improve the learning level of users , Will inevitably continue to introduce users are not familiar with 、 A difficult problem 、 Puzzling video , It is easy to be rejected by users and lead to loss of users . And if the recommended topic video content is very simple for users , Although users have high short-term acceptance , But spending money doesn't buy the improvement of learning level , It leads to the loss of users and also affects the reputation . therefore , The resources launched by the recommendation system must just be able to step out of the user's comfort zone , meanwhile , Users use recommended resources with less effort .

The purpose of this article , On the one hand, introduce the education recommendation system , On the one hand, it also dispels some misunderstandings of many e-commerce mobile Internet practitioners about the education recommendation system , After all, the recommendation system of e-commerce has been developed for many years, and the theory and technology have been relatively complete and mature . However, education recommendation system is closely related to people's deep-seated needs , The educational theory in the information age is still in the exploratory period , In addition to learning from some ideas of the traditional recommendation system , We also need to learn some experience from the education industry , It involves the humanities of the education industry 、 Psychological knowledge is difficult to model mathematically .so, Go and explore .

---

---

---

Reference material ：
Talk about online education recommendation system