=Paper=
{{Paper
|id=Vol-1388/DeCat2015-paper5
|storemode=property
|title=Modeling Short-Term Preferences in Time-Aware Recommender Systems
|pdfUrl=https://ceur-ws.org/Vol-1388/DeCat2015-paper5.pdf
|volume=Vol-1388
|dblpUrl=https://dblp.org/rec/conf/um/BasileCGLS15
}}
==Modeling Short-Term Preferences in Time-Aware Recommender Systems==
https://ceur-ws.org/Vol-1388/DeCat2015-paper5.pdf
Modeling Short-Term Preferences in
Time-Aware Recommender Systems
Pierpaolo Basile, Annalina Caputo, Marco de Gemmis, Pasquale Lops, and
Giovanni Semeraro
Department of Computer Science, University of Bari Aldo Moro,
Via E. Orabona 4, 70125 Bari, Italy
{name.surname}@uniba.it
http://www.di.uniba.it
Abstract. Recommender Systems suggest items that are likely to be
the most interesting for users, based on the feedback, i.e. ratings, they
provided on items already experienced in the past. Time-aware Recom-
mender Systems (TARS) focus on temporal context of ratings in order to
track the evolution of user preferences and to adapt suggestions accord-
ingly. In fact, some people’s interests tend to persist for a long time, while
others change more quickly, because they might be related to volatile in-
formation needs. In this paper, we focus on the problem of building an
effective profile for short-term preferences. A simple approach is to learn
the short-term model from the most recent ratings, discarding older data.
It is based on the assumption that the more recent the data is, the more
it contributes to find items the user will shortly be interested in. We
propose an improvement of this classical model, which tracks the evolu-
tion of user interests by exploiting the content of the items, besides time
information on ratings. When a new item-rating pair comes, the replace-
ment of an older one is performed by taking into account both a decay
function for user interests and content similarity between items, com-
puted by distributional semantics models. Experimental results confirm
the effectiveness of the proposed approach.
Keywords: Time-aware Recommender Systems, Content-based Filter-
ing, Short-Term Preferences, Distributional Semantic Models
1 Introduction
Recommender systems adopts information filtering algorithms to suggest items
or information that might be interesting to users. In general, these systems
analyze the past behavior of a user, build a model or profile of her interests,
and exploit that profile to find potentially interesting items. In collaborative
approaches, the user profile is usually the vector of ratings assigned to all items
that they have accessed, viewed, or purchased [12]. Content-based approaches
rely on item and user descriptions (content) to build item representations and
user profiles that suggest items similar to those a target user already rated (and
liked) in the past [17].
One limitation of these traditional approaches is that the temporal context
of ratings is not taken in account, but actually user preferences are likely to
change over time: long term interests stay stable for a long time, short term
preferences tend to vary with higher frequency. There are some domains, such
as news recommendation, in which retaining this temporal distinction among
user preferences has an impact on the accuracy of suggestions [7, 4]. Indeed, the
issue of including time information into user modeling and recommendation ap-
proaches has been investigated early in literature [3, 22], but the topic recently
received renewed attention, due to the significant improvements of recommen-
dation accuracy obtained by the time-aware algorithm adopted by the winning
team of Netflix Prize competition [15].
Time-aware Recommender Systems (TARS) fall in the more general category
of contex-aware ones, that exploit the context in which users express their prefer-
ences (such as: location, time, weather, emotional state) in order to adapt rating
prediction depending on the situation in which they are experiencing an item.
TARS focus on temporal context of ratings to adapt the recommendation list ac-
cordingly. Regarding the usage of time information, the literature distinguishes
two classes of methods [6]:
– time-aware approaches, that adapt rating predictions on the target recom-
mendation time;
– time-adaptive approaches, which do not differentiate rating predictions ac-
cording to the target time, but rather adjust some parameters or data dy-
namically.
We focus on time-adaptive approaches, specifically on those adopting some
heuristics to penalize older preferences that are presumed to be less valid at
recommendation time. These methods could be considered as a particular case
of time decay heuristics, but they do not target a specific recommendation time
(morning, week-end, etc.). Our investigation specifically concerns approaches
that adopt some time-based strategy to learn separate models for short-term
preferences and for interests that persists for a long time.
In particular, in this paper, we face with the problem of building an effective
short-term preference model able to predict items that will shortly be consumed
by the user. To address this issue, a simple and popular approach is the use
of sliding windows, that learns the model by including in the training set only
the most recent ratings, while older data are discarded or weighted so that they
contribute to the model in a limited way. The literature reports controversial
results about the adoption of time weights for ratings provided at different times.
For example, in [10] the authors show that recommendation accuracy is improved
by an exponential decay function, while an opposite conclusion is drawn by
performing rating prediction on the Netflix Prize dataset [14].
We argue that recent ratings of users reflect their short-term preferences
more than old ratings, but this is not true for all old ratings, but only for
those provided on items which are different from the recently rated ones. In other
words, the hypothesis is that older ratings do not correspond definitely to older
preferences, as assumed in the classical sliding window approach, but content of
the items should be taken into account as well, in order to discard old interests
which differ from new ones.
The research question we want to investigate in this work is: “Is the gradual
decay of influence of ratings, combined with content similarity between items,
useful for modeling short-term preferences?”
We propose a time-aware distributional content-based recommender system
which suggests items the user will shortly be interested in. It exploits both an
exponential decay function and semantic similarity between item descriptions
for regulating the item participation in building of the user profile. Items are
described in a WordSpace through the geometrical metaphora of meanings. Re-
lated words are represented as near points (i.e. vectors), while the semantics
of item descriptions (i.e. text fragments) is computed by summing the vectors
associated with their words.
The paper is organized as follows: in Section 2 we discuss some relevant
literature and compare existing approaches to the proposed one, described in the
following section. Section 4 analyzes the results of the experiments performed to
validate our proposal, while conclusions are drawn in the last section.
2 Related Work
Following the characterization given by [6], we would place our method in the
class of time-adaptive heuristic-based approaches. Rather than modeling the no-
tion of time as a context with respect to the items may be relevant or not,
we consider time as a continuos context attribute with respect to items could
be either fresh or not. Simply stated, our time-context definition aims at mod-
elling the recency of user’s preferences. This point of view has been inspired by
early works by Ding et al. [10, 11], that aimed at adapting collaborative filtering
algorithms in order to capture preference drift.
In [10], the authors propose a novel item-based algorithm (that identifies the
similarity between two items by comparing users’ ratings on them), enhanced
by a time-decay function in a way that the items rated recently contribute more
to the prediction of the recommendations. The underlying assumption is the
same as in the sliding window approach: latest ratings reveal latest interests.
The proposed approach showed the actual improvement of precision obtained
by using an exponential decay function to weight ratings. In the later work [11],
a recency-based approach is proposed, in which each rating of the target user
is assigned a weight that is computed according to its deviation from her most
recent ratings on similar items. We ground our approach on similar basis, but
a significant difference is that we adopt a decay function to weigh content simi-
larity among items recently rated and older ones, in order to select those to be
included in the training set for our model. Time-adaptive heuristics have been
used also by Cao [8] et al. and Lathia et al.[16]. The former approach introduces
four types of user interest patterns and proposes an effective approach for de-
tecting these patterns by exploiting user rating graphs and rating chains. The
authors show that recommendation quality improves when the derived interest
patterns are taken into account. The latter formalises collaborative filtering as a
time-dependent, iterative prediction problem over a dynamic, growing dataset.
The authors propose adaptive temporal collaborative filtering, a method of tem-
porally adapting the size of user kNN neighbourhoods based on the performance
measured up to the current time.
Another method that adopts simple heuristics to improve collaborative filter-
ing is proposed in [5], where the authors describe a time-biased kNN algorithm
exploiting only the most recent ratings from the neighbours. It showed better
performance than other kNN recommendation strategies.
Other approaches adapted factorization algorithms to face with temporal
effects. Remarkable examples of these methods are described in [14] and [21].
Koren [14] suggested that a mere decay of older instances or usage of separate
models for tracking the evolution of preferences cause a loss of prediction accu-
racy. The proposed solution is to model the temporal dynamics along the whole
time period, allowing to separate volatile factors from durable ones, in order to
capture the way user and product characteristics change over time. Xiong et al.
[21] proposed a factorization method based on probabilistic latent factor mod-
els. In addition to the factors that are used to characterize entities, the authors
introduce another set of latent features for each different time period. These
additional factors represent the preference of latent features at each particular
time, so that they are able to capture the evolution of preferences.
Compared to these work which handled the temporal dynamics in different
ways, we focus on explicitly modeling short-term preferences and their influence
on recommendation of items that will shortly be consumed.
Among approaches that propose different models for long-term and short-
term preferences, Cantador et al. [7] designed a content-based news recommender
system in which short-term preferences are inferred from the click history, and
final ranking of items is adapted to the current context of interest. In [20], the
authors propose propose a graph-based approach that introduces session nodes,
associated with a user at specific time, to capture short-term linkages between
items. If two items are connected by session nodes, their similarity is assumed
to be contributed by short-term preferences. An algorithm for preference fusion
is designed for temporal recommendation, that proved to be effective on real
datasets.
Differently from previous methods, our short-term model tries to capture se-
mantic similarity among items by looking at their content, rather than simply
co-occurence within sessions or clicking history, with the hope that the semantic
approach, combined with time information, helps to discover short-term related-
ness among them.
3 Time-Aware Distributional Recommender System
Models of Distributional Semantics have drawn a lot of attention in recent years
due to their capability of capturing semantics at a latent level, without the re-
quirement of learning algorithms or human-edited resources. Such models build
a vector space of meanings where concepts are represented through vectors and
the relatedness between meanings is expressed through a proximity function of
the points they are represented by. Usually these models are built by skimming
a large corpus in order to gather information about distribution of words in a
text. Indeed, statistics about word co-occurrences are useful to infer paradig-
matic relationships among words, i.e. relations about words that can be used
interchangeably. One of the commonest use of such a model is for computing
the similarity between words, since the vector components grasp the semantic
of word usage in context. Then, the vector addition between words belonging to
a text is an easy way to extend to a whole sentence/paragraph/document such
a similarity. This work exploits the idea of adapting Distributional Semantic
Model (DSM) to item descriptions as a unified framework for both representing
the semantic content of items and computing the similarity between them.
3.1 DSM-based Recommender System
The distributional semantic-based recommender system relies on Random In-
dexing (RI) [13] for building up the semantic space. Given a text corpus, RI
technique consists of the following two steps:
1. A random vector is assigned to a term in the corpus vocabulary. This vector
is highly dimensional, with very few elements -called seed - that take values
in {−1, 1}. The dimension of the reduced space corresponds to the random
vector dimensionality;
2. The semantic vector representation for the term is built by analyzing the
whole corpus and summing the random vectors of co-occurring terms in a
given text window.
Mathematically, the sum over random vectors corresponds to multiply the orig-
inal co-occurrence matrix by a projection operator, which preserves the dis-
tance proportion between points. The resulting space, called WordSpace has
two strenght points: 1) the reduced dimension enables a quicker computation
of similarities, and 2) likewise Latent Semantic Analysis, shrinking the number
of components to a smaller set of contexts makes high order relationships more
prominent.
The item space is built upon the previously computed WordSpace: the item
representation comes from the sum of semantic vectors associated to the item
textual content. The user profile, in turn, is built on the basis of the item vector
representations she liked or disliked before. In particular, the model keeps trace
of both positively and negatively rated items and builds two different profile
vectors, u+ and u− , as the sum of positive and negative items, respectively.
In order to get a single profile vector on which basis the model will compute
the recommendations, we exploit the orthogonal projection operator, which has
been successfully employed in both retrieval and recommendation scenarios [2,
18]. The idea behind the use of orthogonalization is that if two vectors are
orthogonal with respect to each other, they do not share components, which
translates into “they have unrelated concepts”. Hence, if we want to express
the user profile through a vector that reflects the positively rated items while
discarding for negative ones, logically we should represent the user vector u as:
− −
u+ + + −
1 ∨ u2 ∨ . . . ∨ un ∧ N OT (u1 ) ∧ N OT (u2 ) ∧ . . . ∧ N OT (um ). However, the
logical aN OT b translates into a vector space endowed with a scalar product as
the projection of a onto the orthogonal space hbi⊥ ≡ {v ∈ V : ∀b ∈ hbi, v · b = 0},
where hbi is the subspace {λb : λ ∈ R}. Thus, computing u corresponds to
summing all items in u+ and then projecting this vector onto the orthogonal
space generated by the vectors in u− . However, following the De Morgan rules,
−
the computation of u can be semplified in u = u+ + +
1 ∨ u2 ∨ . . . ∨ un ∧ N OT (u1 ∨
− −
u2 ∨. . .∨um ), which corresponds to the orthogonalization of two vector (the sum
of positive and negative items) performed through the Gram-Schmidt method.
Then, the model of recommendation consists in exploring the set of non rated
items, in order to assess their similarities with respect to the computed user
profile vector. Such a similarity is computed as the cosine similarity, then the
ranked list of recommended items can be presented to the user.
3.2 Time-adaptive algorithm
A time-adaptive user profile should be able to grasp changes in the user’s be-
haviour in order to reflect the latest user tastes. We tackle this problem by
profiling the user preferences with respect to a sliding window of time: i.e. the
items which contribute to building the profile are those occurred shortly before
the recommendation. This kind of short-term model was initially proposed by
Billsus and Pazzani [3]. However, the user may occasionally manifest a burst of
interest towards new items, then by giving more prominency only to the latest
voted items can result in suggestions that diverge from the real user’s prefer-
ences. The time-adaptive algorithm we propose aims to reflect the recency of
user’s interests in the recommendation process without completely neglecting
the role of items that belong to the remote history of user. Indeed, the set of
items (profile set) which contribute to building the user’s profile is collected by
taking into account two factors:
Time: Recent items contribute more to the profile;
Similarity: The profile set tends to preserve the items whose content is similar
to the newly added one.
Let I = i1 , ..., ik be such a profile set, every time the user rates a new item inew ,
this is automatically added to I, while an older item is removed. The element
to be discarded is selected as follows:
iold = argmin{i ∈ I, sim(i, inew ) · e−λ·(tnew −ti ) } (1)
where t identifies the rating time, for both the new item and that under ob-
servation, sim is the similarity between two items computed in the item space
described in the previous subsection, and λ is a decay factor. Equation 1 aims
to eliminate the item most dissimilar from the newly introduced one. However,
in doing so we try to keep coherent the user profile by weighting this factor with
the exponential function, whose role is to smooth similarity through time. Then,
two possibilities may occur:
1. The new item completely diverges from the user history. In this case all
items in I will take on a small similarity and the exponential function will
contribute more to equation 1;
2. The new item has some degree of similarity with some items in I. In this
case the similarities will be reduced by the exponential function which serves
to mediate the contribute of those similar items rated a long time ago.
Among these, the first scenario is the most interesting, since it reflects a potential
new trend in the user’s preference. Under this condition, the effect of equation
1 on a new incoming item would be that of consolidating this new trend in
the user profile, if a newly added item is similar to that latest one, or quickly
discard the “exception”, thank to the contribute of the exponential function.
The λ parameter plays in this context an important role, since it regulates how
fast the exponential function has to reduce its rate [10].
4 Evaluation
The goal of the evaluation is to assess the capability of the proposed time-
adaptive algorithm to reflect the recency of user interests without loosing infor-
mation about consolidated long-lasting preferences. Then, we compare our pro-
posed algorithm (sim) with respect to a simple sliding window strategy (fifo).
In fifo, the window of items is kept constant: as a new item is added, the older
one is removed, thus following a first-in first-out strategy. This approach dis-
cards items on the basis of a mere time factor, and no information about user
preferences in long-time is preserved.
4.1 Dataset and system setup
The evaluation is performed on the same dataset proposed by Adomavicius et
al. [1] This dataset was originally designed for context-aware recommendation.
However, since it contains rating timestamps, it suits our case. This dataset
comprises 1,757 ratings from 117 users about 226 movies. However, after the
removal of ratings without timestamps, we obtained a total of 1,492 ratings
from 51 users on about 218 movies. The WordSpace is built by collecting co-
occurrences information from a dummy corpus consisting in:
– BNC, a collection of documents from the British National Corpus (BNC)1 ,
containing 100 million words;
1
http://www.natcorp.ox.ac.uk/
– CMU Movie Summary Corpus2 , a dataset of 42,306 movie plot sum-
maries extracted from Wikipedia;
– PLOT, a collection of plots of movies in the Adomavicious corpus, extracted
from Wikipedia.
The objective behind the use of such a corpus is to provide as wider as
possible coverage of all word usages in a language. The WordSpace represents
the top 150, 000 most frequent keywors, the vector dimension is set up to 400,
the number of seeds is of 10 elements, while the window size for computing co-
occurrences is of 5 words. The recommender system is implemented in Java, and
relies on Lucene API3 for building both the WordSpace and the recommendation
model. The factor λ in equation 1 is set to 0.01.
4.2 Evaluation protocol
We evaluate the proposed model on a top-N recommendation task. The eval-
uation is performed as a time-dependent cross-validation, based on increasing
time window [6]. This means that the dataset is split on the basis of temporal
order of rated items. For each user, we order ratings on the temporal line, then
we choose the first k1 elements as training and the following k2 items as test
set, where N has to be chosen ≤ k2 . Then, this methods computes iteratively
the training and test set by adding the previous test to the current training
set, while the new training set is made by sliding k2 items along the temporal
axis. Each iteration corresponds to a single fold. Then, the evaluation metrics
are computed over all the users. We compare the sim approach with respect
to the fifo baseline in terms of Mean Average Precision (MAP) [9] and NDCG
(Normalized Discounted Cumulative Gain) [19], two metrics that are particu-
larly suitable for our evaluation since they take into account the order of items
in the final rank. In fact, the goal of the evaluation is to assess the ability of the
proposed method to suggests items which will shortly be consumed by the user.
Therefore, the ranking computed by our recommendation method is compared
to the ideal ranking defined in the test set by ratings and corresponding time
information. In other words, we want to evaluate whether our method is able to
rank in the top positions of the recommendation list those items in the test set
having high ratings and next to the training set, along the temporal axis.
4.3 Analysis of the Results
We compared the two time-adaptive methods on a variable k1 , i.e. the training set
dimension. We set k2 = 5 and N = 3, while varing k1 ∈ {5, . . . , 15}. We decided
to use this strategy in order to assess the validity of the proposed method when
a wider profile set is available.
Tables 1a and 1b report the results of the evaluation with respect to MAP and
NDCG metrics. Both tables show a similar trend. The best result is obtained on
2
http://www.ark.cs.cmu.edu/personas/
3
http://lucene.apache.org/
k1 = 7 by sim method, which gives also better overall figures. There are only two
exceptions to this trend, with k1 = 11 and k1 = 14, although the differences in
reported values are very small. Another common trend showed by results is that,
as we increase the k1 dimension, the differences between sim and fifo become
smaller. However, we ascribe such a trend to the fact that by increasing the
dimension of the training set, the differences between fifo and sim strategies
become smaller since the variability in the set decreases (i.e. an increasingly
number of items from the past of the user become part of the training set).
Table 1: Evaluation results at different size of k1 .
(a) MAP (b) NDCG.
k1 fifo sim k1 fifo sim
5 0.447 0.460 5 0.555 0.569
6 0.444 0.469 6 0.552 0.575
7 0.465 0.499 7 0.569 0.597
8 0.456 0.461 8 0.559 0.564
9 0.459 0.469 9 0.560 0.575
10 0.457 0.466 10 0.566 0.573
11 0.458 0.455 11 0.562 0.557
12 0.459 0.468 12 0.558 0.565
13 0.449 0.457 13 0.550 0.557
14 0.460 0.458 14 0.567 0.564
15 0.431 0.451 15 0.537 0.555
5 Conclusion and Future Work
Generally, the behavior of a user may be determined by her long-term inter-
ests, but at any given time, she is also affected by short-term preferences or
information needs. In this paper, we argue that time is not the only factor to
be taken into account in order to distinguish among short-time and long-time
interests. We started from the classical sliding window approach and suggested
that older ratings do not correspond definitely to older preferences, but content
of the items should be considered as well. We proposed an approach that models
short-term preferences by adopting a content-based sliding window approach:
when a new ratings comes into the system, the replacement of an older one is
performed by taking into account both a decay function for user interests and
content similarity between items on which ratings are provided, computed by
distributional semantics models. We compared the proposed approach to the
simple FIFO strategy (the new rating replaces the oldest one). Experimental
results confirmed the hypothesis that the gradual decay of influence of ratings,
combined with content similarity between items, is actually useful for model-
ing short-term preferences, especially when a few ratings are available to train
the system. As a future work, we plan to evaluate our short-term model on a
wider dataset. Furthermore, we want to design a model for long-term prefer-
ences, as well as a way to integrate the two models in order to have an overall
recommendation strategy.
Acknowledgements
This work fulfils the research objectives of the PON 02 00323 2938699 CUP
B86D130000300007 project “EFFEDIL - SOLUZIONI INNOVATIVE PER
L’EFFICIENZA ENERGETICA IN EDILIZIA” funded by the Italian Min-
istry of University and Research (MIUR).
References
1. Adomavicius, G., Sankaranarayanan, R., Sen, S., Tuzhilin, A.: Incorporating con-
textual information in recommender systems using a multidimensional approach.
ACM Trans. Inf. Syst. 23(1), 103–145 (Jan 2005)
2. Basile, P., Caputo, A., Semeraro, G.: Negation for document re-ranking in ad-
hoc retrieval. In: Amati, G., Crestani, F. (eds.) Advances in Information Retrieval
Theory, Lecture Notes in Computer Science, vol. 6931, pp. 285–296. Springer Berlin
Heidelberg (2011)
3. Billsus, D., Pazzani, M.J.: User modeling for adaptive news access. User Model.
User-Adapt. Interact. 10(2-3), 147–180 (2000), http://dx.doi.org/10.1023/A:
1026501525781
4. Billsus, D., Pazzani, M.J.: The adaptive web. chap. Adaptive News Access, pp. 550–
570. Springer-Verlag, Berlin, Heidelberg (2007), http://dl.acm.org/citation.
cfm?id=1768197.1768218
5. Campos, P.G., Bellogı́n, A., Dı́ez, F., Chavarriaga, J.E.: Simple time-biased knn-
based recommendations. In: Proceedings of the Workshop on Context-Aware Movie
Recommendation. pp. 20–23. CAMRa ’10, ACM, New York, NY, USA (2010)
6. Campos, P., Dez, F., Cantador, I.: Time-aware recommender systems: a compre-
hensive survey and analysis of existing evaluation protocols. User Modeling and
User-Adapted Interaction 24(1-2), 67–119 (2014), http://dx.doi.org/10.1007/
s11257-012-9136-x
7. Cantador, I., Bellogı́n, A., Castells, P.: Ontology-based personalised and
context-aware recommendations of news items. In: Proceedings of the 2008
IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent
Agent Technology - Volume 01. pp. 562–565. WI-IAT ’08, IEEE Computer Soci-
ety, Washington, DC, USA (2008), http://dx.doi.org/10.1109/WIIAT.2008.204
8. Cao, H., Chen, E., Yang, J., Xiong, H.: Enhancing recommender systems under
volatile userinterest drifts. In: Proceedings of the 18th ACM Conference on Infor-
mation and Knowledge Management. pp. 1257–1266. CIKM ’09, ACM, New York,
NY, USA (2009), http://doi.acm.org/10.1145/1645953.1646112
9. Croft, B., Metzler, D., Strohman, T.: Search Engines: Information Retrieval in
Practice. Addison-Wesley Publishing Company, USA, 1st edn. (2009)
10. Ding, Y., Li, X.: Time weight collaborative filtering. In: Proceedings of the 14th
ACM International Conference on Information and Knowledge Management. pp.
485–492. CIKM ’05, ACM, New York, NY, USA (2005), http://doi.acm.org/10.
1145/1099554.1099689
11. Ding, Y., Li, X., Orlowska, M.E.: Recency-based collaborative filtering. In: Pro-
ceedings of the 17th Australasian Database Conference - Volume 49. pp. 99–107.
ADC ’06, Australian Computer Society, Inc., Darlinghurst, Australia, Australia
(2006), http://dl.acm.org/citation.cfm?id=1151736.1151747
12. Herlocker, J.L., Konstan, J.A., Borchers, A., Riedl, J.: An algorithmic framework
for performing collaborative filtering. In: SIGIR ’99: Proceedings of the 22nd An-
nual International ACM SIGIR Conference on Research and Development in In-
formation Retrieval, August 15-19, 1999, Berkeley, CA, USA. pp. 230–237. ACM
(1999), http://doi.acm.org/10.1145/312624.312682
13. Kanerva, P.: Sparse Distributed Memory. MIT Press (1988)
14. Koren, Y.: Collaborative filtering with temporal dynamics. In: Proceedings of the
15th ACM SIGKDD International Conference on Knowledge Discovery and Data
Mining. pp. 447–456. KDD ’09, ACM, New York, NY, USA (2009), http://doi.
acm.org/10.1145/1557019.1557072
15. Koren, Y.: Collaborative filtering with temporal dynamics. Commun. ACM 53(4),
89–97 (Apr 2010), http://doi.acm.org/10.1145/1721654.1721677
16. Lathia, N., Hailes, S., Capra, L.: Temporal collaborative filtering with adaptive
neighbourhoods. In: Proceedings of the 32Nd International ACM SIGIR Confer-
ence on Research and Development in Information Retrieval. pp. 796–797. SIGIR
’09, ACM, New York, NY, USA (2009)
17. Lops, P., de Gemmis, M., Semeraro, G.: Content-based recommender systems:
State of the art and trends. In: Ricci, F., Rokach, L., Shapira, B., Kantor, P.B.
(eds.) Recommender Systems Handbook, pp. 73–105. Springer (2011), http://dx.
doi.org/10.1007/978-0-387-85820-3_3
18. Musto, C., Semeraro, G., Lops, P., de Gemmis, M.: Random indexing and negative
user preferences for enhancing content-based recommender systems. In: Huemer,
C., Setzer, T. (eds.) E-Commerce and Web Technologies, Lecture Notes in Business
Information Processing, vol. 85, pp. 270–281. Springer Berlin Heidelberg (2011)
19. Shani, G., Gunawardana, A.: Evaluating recommendation systems. In: Ricci, F.,
Rokach, L., Shapira, B., Kantor, P.B. (eds.) Recommender Systems Handbook, pp.
257–297. Springer US (2011)
20. Xiang, L., Yuan, Q., Zhao, S., Chen, L., Zhang, X., Yang, Q., Sun, J.: Temporal
recommendation on graphs via long- and short-term preference fusion. In: Proceed-
ings of the 16th ACM SIGKDD International Conference on Knowledge Discovery
and Data Mining. pp. 723–732. KDD ’10, ACM, New York, NY, USA (2010),
http://doi.acm.org/10.1145/1835804.1835896
21. Xiong, L., Chen, X., Huang, T., Schneider, J.G., Carbonell, J.G.: Temporal col-
laborative filtering with bayesian probabilistic tensor factorization. In: Proceed-
ings of the SIAM International Conference on Data Mining, SDM 2010, April
29 - May 1, 2010, Columbus, Ohio, USA. pp. 211–222. SIAM (2010), http:
//dx.doi.org/10.1137/1.9781611972801.19
22. Zimdars, A., Chickering, D.M., Meek, C.: Using temporal data for making recom-
mendations. In: Proceedings of the 17th Conference in Uncertainty in Artificial
Intelligence. pp. 580–588. UAI ’01, Morgan Kaufmann Publishers Inc., San Fran-
cisco, CA, USA (2001), http://dl.acm.org/citation.cfm?id=647235.720264