Time-LSTM：提升用户行为建模的推荐系统

24 浏览量更新于2024-08-26 收藏 294KB PDF 举报

"通过Time-LSTM对用户行为进行建模的研究论文" 在推荐系统（RS）领域，递归神经网络（RNN），特别是长短期记忆网络（LSTM），已经成为一种流行的解决方案。这是因为用户的行为序列往往展现出一定的模式，而RNN在处理顺序数据时表现突出。然而，传统的RNN在建模时往往只关注对象的顺序，忽略了时间间隔这一关键因素。在推荐系统中，用户行为之间的时序关系对于理解用户兴趣和行为模式至关重要。 Time-LSTM是针对这一问题提出的一种新型LSTM变体，它引入了时间门机制来处理时间间隔。这些时间门的设计目的是增强模型对用户短期和长期兴趣的捕获能力，从而提升推荐的准确性。通过这种方式，Time-LSTM能够更有效地模拟用户在不同时间点的行为变化，进一步优化推荐性能。论文的作者来自中国和加拿大的多所高校，他们在实验部分展示了Time-LSTM在两个真实数据集上的表现，结果表明Time-LSTM推荐方法相比于传统方法有显著的改进。这不仅验证了时间间隔在用户行为建模中的重要性，也证实了Time-LSTM架构的有效性。在实际应用中，这种时间感知的建模方法可以为推荐系统提供更准确的用户行为预测，从而实现更个性化的推荐。例如，考虑到用户在购买书籍后的阅读速度，或者观看电影后的评论间隔，Time-LSTM可以更精确地推断用户的下一步可能行为，从而提升用户体验和系统整体的推荐质量。这篇研究论文深入探讨了在推荐系统中如何利用RNN，尤其是Time-LSTM，来建模和理解用户行为的时序模式。通过时间门的引入，Time-LSTM在捕获用户短期和长期兴趣方面表现出优越性，对于提升推荐系统的性能具有重要的理论和实践意义。

What to Do Next: Modeling User Behaviors by Time-LSTM

Yu Zhu

†

, Hao Li

†

, Yikang Liao

†

, Beidou Wang

]‡

, Ziyu Guan

, Haifeng Liu

]

, Deng Cai

†∗

†

State Key Lab of CAD&CG, College of Computer Science, Zhejiang University, China

College of Information and Technology, Northwest University of China

]

College of Computer Science, Zhejiang University, China

‡

School of Computing Science, Simon Fraser University, Canada

{zhuyu cad, haolics, ykliao, haifengliu, dcai}@zju.edu.cn, beidouw@sfu.ca, ziyuguan@nwu.edu.cn

Abstract

Recently, Recurrent Neural Network (RNN) solu-

tions for recommender systems (RS) are becoming

increasingly popular. The insight is that, there ex-

ist some intrinsic patterns in the sequence of users’

actions, and RNN has been proved to perform ex-

cellently when modeling sequential data. In tradi-

tional tasks such as language modeling, RNN so-

lutions usually only consider the sequential order

of objects without the notion of interval. However,

in RS, time intervals between users’ actions are of

signiﬁcant importance in capturing the relations of

users’ actions and the traditional RNN architectures

are not good at modeling them. In this paper, we

propose a new LSTM variant, i.e. Time-LSTM,

to model users’ sequential actions. Time-LSTM

equips LSTM with time gates to model time inter-

vals. These time gates are speciﬁcally designed,

so that compared to the traditional RNN solutions,

Time-LSTM better captures both of users’ short-

term and long-term interests, so as to improve the

recommendation performance. Experimental re-

sults on two real-world datasets show the superi-

ority of the recommendation method using Time-

LSTM over the traditional methods.

1 Introduction

Recurrent Neural Network (RNN) solutions have become

state-of-the-art methods on modeling sequential data. They

are applied to a variety of domains, ranging from language

modeling to machine translation to image captioning. With

remarkable success achieved when RNN is applied to afore-

mentioned domains, there is an increasing number of works

trying to ﬁnd RNN solutions in the area of recommender sys-

tems (RS).

[

Hidasi et al., 2016a; Tan et al., 2016; Hidasi et al.,

2016b

]

focus on RNN solutions in one certain type of recom-

mendation task, i.e. session-based recommendations, where

no user id exists and recommendations are based on previous

consumed items within the same session.

[

Yu et al., 2016

]

points out that RNN is able to capture users’ general inter-

est and sequential actions in RS and designs a RNN method

∗

corresponding author

(a) Language Modeling

m+1

△t

(b) Recommender Systems

Figure 1: w

in (a) represents the m-th word. In (b), i

represents

the m-th consumed item and 4t

is the time interval between the

time when i

and i

m+1

are consumed.

for the next-basket recommendations. The insight that RNN

works well in the above recommendation tasks is that, there

exist some intrinsic patterns in the sequence of users’ actions,

e.g. once a man buys a badminton racket, he tends to buy

some badmintons later, and RNN has been proved to perform

excellently when modeling this type of patterns.

However, none of the above RNN solutions in RS consid-

ers the time interval between users’ neighbour actions, while

these time intervals are important to capture the relations of

users’ actions, e.g. two actions within a short time tend to be

related and actions with a large time interval may aim at dif-

ferent goals. Therefore, it is important to exploit the time in-

formation when modeling users’ behaviors, so as to improve

the recommendation performance. We use Figure 1 to show

what the time interval is and how it makes RS different from

the traditional domains such as language modeling. Speciﬁ-

cally, there is no notion of interval between neighbour words

(e.g. no interval between w

and w

) in language modeling,

while there are time intervals between neighbor actions (e.g.

between i

and i

) in RS. Traditional RNN architectures

are good at modeling the order information of sequential data

as in Figure 1 (a), but they cannot well model time intervals

in Figure 1 (b). Therefore, new models need to be proposed

to address this problem.

A recently proposed model, i.e. Phased LSTM

[

Neil et al.,

2016

]

, tries to model the time information by adding one time

gate to LSTM

[

Hochreiter and Schmidhuber, 1997

]

, where

LSTM is an important ingredient of RNN architectures. In

this model, the timestamp is the input of the time gate which

controls the update of the cell state, the hidden state and

thus the ﬁnal output. Meanwhile, only samples lying in the

model’s active state are utilized, resulting in sparse updates

during training. Thus, Phased LSTM can obtain a rather fast

Proceedings of the Twenty-Sixth International Joint Conference on Artiﬁcial Intelligence (IJCAI-17)

3602

下载后可阅读完整内容，剩余6页未读，立即下载

weixin_38701156

粉丝: 5
资源: 957

Time-LSTM：提升用户行为建模的推荐系统

Time-RNNs:时间LSTM的Pytorch实现，如“下一步做什么”中所示

LSTM中timesteps的理解

淘宝用户行为数据集

考虑用户意图和时间间隔的会话型深度学习推荐系统.pdf

【股市预测实战】：LSTM模型的理论基础与应用演练

【GRU与Peephole深度分析】：LSTM变体的创新与挑战

【LSTM详解】：循环神经网络的变体深度剖析

LSTM案例实战：时间序列分析中的顶尖表现与深度解读

【图像识别中的LSTM】：探索前沿应用，技术实践揭秘

MATLAB回归分析：时间序列数据的建模与预测实战策略

最新资源