Recommended decision comparison user-based and item-based recommendation algorithms

Source: Internet
Author: User

Transferred from: http://blog.csdn.net/hxxiaopei/article/details/7695740

Recommended system, overall there are three ways:
    • User->user->item, recommended for users with the same interests as item,user-based
    • User->item->item, recommend a item,item-based similar to the item you like
    • User->item-feature->item, extract the characteristics of the user like item, recommend the item with these characteristics, model-based
  for the top two, most of the current use is item-based, such as Amazon, talk about Item-base better than user-based, but similar to Digg, or use user-based, the effect is also good, So not one must be better than the other, just use a different scene.   Compare these two methods: user-based more Consider the same hobby of the user interest, recommend these users like/access to the item, and the user's current behavior is not much, more is the user of these friends visited what, belong to the social behavior of the circle, The recommended item is the same hobby user favorite item, so has a hot spot effect, that is, the recommended circle user access to the most, but also can be the circle users just access to the item recommended, with strong real-time, especially the newly introduced hot spots, can quickly spread, can also solve the new-item cold start problem. Item-based mainly consider the user's historical interests, recommend and user history like item similar to the item, and the user's current behavior has a great relationship between the recommended item and the user's current click Similarity, the user is understandable, that is, the so-called interpretative is very strong, The recommended item is also not popular, it is likely to be unpopular (long tail), but related to the user's interest, require users on this site interest is long and fixed, the recommendation is to help users find and their interests related to the item. Recommended item and which user relationship is not big, so better to solve the problem of new users.   Therefore, when making recommendation decisions, the following questions are considered: 1. Whether the user has a fixed and long-term interest in the field 2. The size of the user's scale is 3.item 4. The speed of the new user 5. The speed of new item 6. Real-time requirements   I think the first one should be the main requirement, reflect the user to the degree of personalization requirements: such as reading and other news hotspot site, and similar to Youku such user-made item site, the user is interested in what happened, what everyone is looking at, and the subdivision of the requirements of the field is not high, users see "Dragon Boat Festival Yantai car crashed bicycle team" video, but also want to see " Youku I am the legendary broken brother inspirational singing "video, and user history may be looking at" Vicky turn "or" chopsticks Brother's father ", there is no obvious theme, also can not reflect the user's interest, the user's current behavior, also do not take the next recommendation immediately effective, Users are only real-time to the popular or other people watching things interested in, while the item new speed is also relatively fast, can quickly let a lot of users to observe, and user growth is slow, this time user-based is a good choice. However, a site like Douban,amazon, the user's current behavior has a significant impact on the recommended item, such as the user's DOuban I read the following to see a recommendation system practice, to demonstrate that users are interested in data mining or recommendation field, then the recommended "Recommend System Handbook" is a good choice, but also very strong interpretation, the recommendation system needs to do is to help the group of users to discover the knowledge of the field , users do not know before, to meet the needs of users. In fact, the analysis of user behavior, the user behavior of focus in these areas, has continuity. In general, the speed at which the website item is increased is not likely to increase frequently, such as new books/Movies. And the new user just click on an item can form a recommendation, so, item size is small, or smaller than the size of the user, while the item change speed is not fast, item-based is the best choice.   Finishing so much, not necessarily correct, slowly feel, change at any time

Recommended decisions Compare user-based and item-based recommended algorithms

Contact Us

The content source of this page is from Internet, which doesn't represent Alibaba Cloud's opinion; products and services mentioned on that page don't have any relationship with Alibaba Cloud. If the content of the page makes you feel confusing, please write us an email, we will handle the problem within 5 days after receiving your email.

If you find any instances of plagiarism from the community, please send an email to: info-contact@alibabacloud.com and provide relevant evidence. A staff member will contact you within 5 working days.

A Free Trial That Lets You Build Big!

Start building with 50+ products and up to 12 months usage for Elastic Compute Service

  • Sales Support

    1 on 1 presale consultation

  • After-Sales Support

    24/7 Technical Support 6 Free Tickets per Quarter Faster Response

  • Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.