大数据分析全英文 (34).pdf

《大数据分析全英文 (34).pdf》由会员分享，可在线阅读，更多相关《大数据分析全英文 (34).pdf（20页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、2Big Data ApplicationComputing AlgorithmComputing ModelData processing systemComputing Platform&EngineData Storing SystemSpark MLlibData VisualizationData Products and Data ServicesBig Data ApplicationData Application systemTensorFlowRecommendation SystemSocial Networking3Recommendation AlgorithmsMa

2、in AlgorithmsNeighborhood-basedUser-based FilteringItem-based Filtering1 Collaborative Filtering2 Content Based Filtering3 Knowledge BasedModel-basedMatrix Factorizationlatent Dirichlet allocation(LDA)Structured FeatureUnstructured FeatureLatent Factor modelGraph model.4Collaborative filtering V.S C

3、ontent based The similar users are defined based on the common articles they read!The similar articles are defined based on the articles features!5User-based filtering V.S Item-based filteringIce cream coneIce cream with Egg RollIce cream coneIce cream with Egg Roll6Four Recommendation System Algori

4、thms User-based filtering Item-based filtering Content-Based Filtering7User-based filtering 1 User purchase records2 Item-User table3 User-User matrix4 User User similarity matrix5 Get users interest in different items6 Get recommendation result8User-based collaborative filtering Similarity calculat

5、ionJaccard similarityCosine similarityUsually use Jaccard similarity,cosine similarity,Euclidean distance andPearson distance to calculate the similarity between two users.Let N(u)be the set of items that user U likes,and N(v)be the set of items thatuser V likes.9User-based collaborative filtering H

6、ow to get similar usersUser purchase recordsItem-User tableUser-User matrix10User-based collaborative filtering How to get similar usersCosine similarityUser-User matrixUser User similarity matrixN(u)is the set of items that user U likes,and N(v)is the set of items that user V likes.11User-based col

7、laborative filtering For each candidate item i,the degree that the user u is interested in it:How to get users interest in different itemsr represents user vs interest in item i,here r=112User-based collaborative filtering How to get recommendation resultUser A likes c and e equallyUser User similar

8、ity matrix13Four Recommendation System Algorithms User-based filtering Item-based filtering Content-Based Filtering14Item-based collaborative filteringCalculate the similarity between items and recommend items with high similarity to the items chosen by users.In the item-based collaborative filterin

9、g algorithm,The similarity between item A and item B is based on the fact that users who like item A also like item B.User purchase recordsItem-Item matrix15Item-based collaborative filteringLet N(i)be the set of users who like item i,and N(j)be the set ofusers who like item j,and|N(i)N(j)|be the nu

10、mber of users wholike both item i and item j.Similarity of item i and item j(cosine similarity):Similarity calculation16Item-based collaborative filteringItem-Item matrixItem-Item similarity matrixSimilarity calculation17Item-based collaborative filteringFor each candidate item j,the degree user u i

11、nterested in it is:N(u)represents the collection of items favored by user U,S(j,k)is the collection of k items most similar to item j,wijis thesimilarity between item i and item j,and ruirepresents user us interestin item i.How to get the users interest in different items18Item-based collaborative f

12、ilteringHow to get recommendation resultFor user B,K=3 similar items,so the items user B hasnot chosen before are b,d and e.then calculate P(B,d),P(B,b)and P(B,e)respectively:User B likes d most and b,e equallyItem-Item similarity matrix19Compare user-based and item-basedUser-based and item-based lo

13、oks like symmetry.user-user similarity matrixPurchase recorditem-item similarity matrix20Four Recommendation System Algorithms User-based filtering Item-based filtering Content-Based Filtering21Content-based recommendationThree main steps:1.Item Representation2.Profile Learning3.Recommendation Generation