模拟国际会议发言稿
- 1、下载文档前请自行甄别文档内容的完整性,平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益,请联系客服反馈,我们会尽快为您处理(人工客服工作时间:9:00-18:30)。
模拟国际会议发言稿
【篇一:模拟国际会议演讲稿】
recsplorer:recommendation algorithms based on precedence mining
1. introduction
thank you very much, dr. li, for your kind introduction. ladies and gentlemen, good morning! i am honored to have been invited to speak at this conference. before i start my speech, let me ask a question. do you think recomemdations from others are useful for your internet shopping? thank you. it is obvious that recommendations play an important role in our daily consumption decisions.
today, my topic is about recommendation algorithms based on precedence mining. i want to share our interesting research result on recommendation algorithms with you. the content of this presentation is divided into 5 parts: in session 1, i will intruduce the tradictional recommendation and our new strategy; in session 2, i will give the formal definition of precedence mining; in session 3, i will talk about the novel recommendation algorithms; experimental result will be showed in session 4; and finally, i will make a conclusion.
2. body
session 1: introduction
the picture on this slide is an instance of recommemdation application on amazon.
recommender systems provide advice on products, movies,web pages, and many other topics, and have become popular in many sites, such as amazon. many systems use collaborative filtering methods. the main process of cf is organized as follow: first, identify users similar to target user; second, recommend items based on the similar users. unfortunately, the order of consumed items is neglect. in our paper, we consider a new recommendation strategy based on precedence patterns. these patterns may encompass user preferences, encode some logical order of options and capture how interests evolve.
precedence mining model estimate the probability of user future consumption based on past behavior. and these
probabilities are used to make recommendations. through our experiment, precedence mining can significantly improve recommendation performance. futhermore, it does not suffer from the sparsity of ratings problem and exploit patterns across all users, not just similar users.
this slide demonstrates the differences between collaborative filtering and precedence mining. suppose that the scenario is about course selection. each quarter/semester a student chooses a course, and rates it from 1 to 5. figure a) shows five transcripts, a transcript means a list of course. u is our target student who need recommendations. figure b) illustrates how
cf work. assume similar users share at least two common courses and have similar rating, then u3 and u4 are similar to u, and their common course h will be a recommendation to u. figure c) presents how precedence mining work. for this example, we consider patterns where one course follows another. suppose patterns occour at least two transcrips are recognized as significant, then (a,d), (e,f) and (g,h) are found out. and d, h, and f are recommendation to u who has taken a, g and e.
now i will a probabilistic framework to solve the precedence mining problems. our target user has selected course a , we want to compute the probability course x will follow, i.e.,
pr[x|a].
﹁howerve, what we really need to calculate is pr[x|ax] rather than pr[x|a]. because in our context,
we are deciding if x is a good recommendation for the target user that has taken a. thus we know that our target user’s transcript does not have x before a. for instance, the transcript no. 5 will be omitted. in more common situation, our target
user has taken a list of courses, t = {a,b,c,…} not
﹁just a. thus, what really need is pr[x|tx]. the question is how to figure out this probability. i will
answer it later.
session 2: precedence mining
we consider a set d of distinct courses. we use lowercase letters (e.g., a, b, … ) to refer to courses in d. a transcript t is a sequence of courses, e.g., a - b - c - d. then the definition of
top-k recommendation problem is as follows. given a set transcripts over d for n users, the extra transcript t of a target