=Paper= {{Paper |id=Vol-2140/paper13 |storemode=property |title=A Collaborative Ranking Model with Contextual Similarities for Venue Suggestion |pdfUrl=https://ceur-ws.org/Vol-2140/paper13.pdf |volume=Vol-2140 |authors=Mohammad Aliannejadi,Fabio Crestani |dblpUrl=https://dblp.org/rec/conf/iir/AliannejadiC18 }} ==A Collaborative Ranking Model with Contextual Similarities for Venue Suggestion== https://ceur-ws.org/Vol-2140/paper13.pdf

A Collaborative Ranking Model with Contextual
Similarities for Venue Suggestion

Mohammad Aliannejadi and Fabio Crestani

Faculty of Informatics, Università della Svizzera italiana (USI), Lugano, Switzerland
{mohammad.alian.nejadi,fabio.crestani}@usi.ch

Abstract. While recent studies have explored the idea of adopting col-
laborative ranking (CR) for recommendation, there has been no attempt
to incorporate contextual similarities between venues. In this study, we
explore the effect of incorporating contextual similarities into the learn-
ing strategy of a CR model. By enhancing the latent associations between
users with contextual similarities, our experiments show that contextual
similarities improve the performance of CR.

1 Introduction

Generating venue suggestions plays a crucial role in satisfying the user needs,
for example when exploring a new city [2]. Recommendation algorithms can be
divided into two categories: content-based and collaborative-based approaches.
Content-based approaches build user and item profiles based on items’ contents
and measure the similarity between the profiles [3,6]. Many real-world problems
limit the accuracy of venue suggestion. For instance, a major issue is the sparsity
of users’ check-in data. To address the data sparsity problem relevant studies
exploit auxiliary information, such as user tags and temporal information [5,9].
Moreover, in relevant literature item recommendation, such as venue suggestion,
is often treated as a rating prediction or matrix completion task [1]. However,
considering the square loss as a measure of prediction effectiveness is not accurate
in the top-N recommendation task [8]. In other words, being able to present a
more accurate ranked list to a user should be rewarded. Collaborative ranking
(CR) is based on this idea and focuses on the accuracy of recommendation at
the top of the list for each user, by learning the individual’s ranking functions
in a collaborative manner [7].
In this study, we explore the effect of incorporating contextual similarities of
venues into CR. We design the objective function of the CR model to include
the contextual similarity measures in the loss function with a focus on ranking
relevant venues at the top of the recommendation list. This enables our model to
propagate venue contextual proximity to the users, thus addressing the sparsity

IIR 2018, May 28-30, 2018, Rome, Italy. Copyright held by the author(s).
problem that appears in the check-in data. For example, in the conventional col-
laborative filtering approaches, the latent associations are captured only if users
have visited the same venue, whereas incorporating venue similarities in our CR
model can capture the associations even if users have only visited contextually
similar venues but not the same ones.

2 Proposed Method

Let P = {ρ1 , . . . , ρn } and L = {l1 , . . . , lm } be the sets of n users and m venues,
respectively. We consider user ratings 1, 2, and 3 on venues as negative feedback,
while ratings 4 and 5 as positive one. For each user ρi , we define L+ i as the set of
relevant venues, and L− i as the set of irrelevant ones. Moreover, let Sz ∈ Rm×m
be the similarity matrix of venues based on a contextual feature z.

Contextual Similarities. In our approach we compute a contextual similarity
between two venues li and lj based on their content and location. In the following
we briefly introduce three similarity measures, defining Sij = {Sz (i, j) : z ∈
{1, 2, 3}} as the set of contextual similarity functions.

– Geographical: first, we compute the geographical similarity between two
venues to incorporate the geographical context while characterizing the user’s
geographical preferences. The similarity is inversely proportional to the dis-
tance between two venues, denoted by S1 (i, j).
– Review based: for venue li , we train a Support Vector Machine (SVM)
classifier with linear kernel to estimate the review-based similarity. We con-
sider positive reviews as positive training samples and negative reviews as
negative training samples to train the SVM and call the trained classifier
SVMi . Then, for each venue lj : j ∈ L we classify the reviews of lj using
SVMi . We take the value of the decision function as the similarity measure,
denoted by S2 (i, j).
– Category based: S3 (i, j) is the cosine similarity between the category vec-
tors of li and lj .

Collaborative Ranking with Contextual Similarities. Here we present
our model, called CRCS, which suggests venues for each user ρi placing relevant
venues at the top of the recommendation list. Our goal is to understand the
user’s check-in behavior with the contextual similarities of venues explained in
Section 2. For example, a user may like all venues that are in the city center and
serve pizza. Building ranking functions considering different contextual similar-
ities between venues also allows us to model latent associations between users
with similar tastes who would not be considered in a traditional CR setting. This
happens because CRCS takes into account the venue similarities as it updates
the user and item latent matrices. CRCS can build the associations between users
as it considers content- and context-based similarities while updating the latent
matrices. Notice that our CRCS model does not rely on the type of contextual
similarity and is not limited to a certain type of contextual features. Hence, it
can be a general framework for incorporating any type of contextual features.
We focus on ranking the venues that a user likes higher than the ones she
does not. Formally, we aim at ranking venues that belong to L+ i higher than
those that are in L− i . Our goal is to rank the venues with emphasis on the top
of the list. Let Hi (lj− ) be the “height” of an irrelevant venue, that is:
3 h
X X i
Hi (lj− ) =

αz × 1[fi (l+ )≤fi (l− )] /Sz (k, j) ,
k j
k∈L+ z=1
i

where αz is the weight of contextual similarity Sz and 1[.] is an indicator func-
tion. Dividing the indicator function by Sz allows the model to incorporate the
contextual similarities into the model while constructing the height for irrele-
vant items. For example, if an irrelevant item is ranked higher than a relevant
item, but they are contextually very similar based on Sz , then the denominator
will be higher, which means the height of the irrelevant venue will be reduced
accordingly. The objective function should aim at minimizing Hi for all irrele-
vant venues of user ρi . A lower value of Hi means that there are fewer irrelevant
venues ranked higher than relevant ones, and those that are ranked higher are
more similar to relevant items. However, indicator functions are not convex and
they are not suitable to our optimization strategy. Therefore, we use the logis-
tic loss of the difference between the two functions as a convex upper bound
surrogate. We define the difference between the k th venue and the j th as follows:
3
X
δi (k, j) = uTi

αz (vk − vj )/ exp(|Sz (k, j)|) .
z=1

Therefore, the surrogate height function Hi0 (lj− ) becomes:
X
Hi0 (lj− ) =

log 1 + exp − δi (k, j) .
k∈L+
i

Finally, the objective function is defined as follows:
m
X 1 X 2
R(U, V ) = Hi0 (lj− ) .
ni
i=1 j∈L−
i

3 Results
We evaluate our approach on a mixed dataset of two benchmark datasets, made
available by the TREC. The datasets are for the TREC Contextual Suggestion
Track (TREC-CS) 2015 and 2016. We used the publicly available crawls of [4].
Table 1 reports the performance of all the models on TREC-CS in terms
of nDCG@k with k ∈ {1, 2, 3, 4, 5}. Our proposed CRCS model significantly
Table 1: Performance evaluation on TREC-CS in terms of nDCG@k with k
∈ {1, 2, 3, 4, 5}. Bold values denote the best scores, for p <0.05 in paired t-test.
nDCG@1 nDCG@2 nDCG@3 nDCG@4 nDCG@5
P-Push 0.5635 0.5282 0.5188 0.4963 0.4775
RH-Push 0.4606 0.4561 0.4581 0.4611 0.4575
IRenMF 0.5037 0.4781 0.4806 0.4759 0.4689
GeoMF 0.4743 0.4842 0.4879 0.4801 0.4774
Rank-GeoFM 0.5662 0.5491 0.5445 0.5123 0.4976
CRCS 0.6832 0.6226 0.5888 0.5522 0.5332

outperforms all state-of-the-art methods in terms of nDCG@k for all values of
k (according to pairwise t-test at p < 0.05). Compared to the state-of-the-art
method, Rank-GeoFM, the improvements in terms nDCG@1 and nDCG@5 are
21% and 7%, respectively. This indicates that our proposed CRCS can address
the data sparsity problem by incorporating different types of contextual similar-
ities. While the geographical similarity includes the neighborhood influences in
the model, the category-based similarity takes into account users with similar
tastes when they do not share the same check-in records. In addition to that,
the review-based similarity models venues similarities in terms of other users’
opinions in various contexts. Fusing these similarity measures with a CR-based
model enables CRCS to form complicated similarity affinities among venues and
propagate it to the users. Hence, our proposed CRCS addresses the data spar-
sity problem better than other state-of-the-art models, indicated by the high
recommendation accuracy.

References
1. G. Adomavicius and A. Tuzhilin. Toward the next generation of recommender
systems: A survey of the state-of-the-art and possible extensions. IEEE Trans.
Knowl. Data Eng., 17(6):734–749, 2005.
2. M. Aliannejadi and F. Crestani. Venue appropriateness prediction for personalized
context-aware venue suggestion. In SIGIR, 2017.
3. M. Aliannejadi and F. Crestani. Personalized context-aware point of interest rec-
ommendation. ACM Trans. Inf. Syst., 2018.
4. M. Aliannejadi, I. Mele, and F. Crestani. A cross-platform collection for contextual
suggestion. In SIGIR, 2017.
5. M. Aliannejadi, D. Rafailidis, and F. Crestani. Personalized keyword boosting for
venue suggestion based on multiple LBSNs. In ECIR, 2017.
6. S. A. Bahrainian, S. M. Bahrainian, M. Salarinasab, and A. Dengel. Implementation
of an intelligent product recommender system in an e-store. In AMT, 2010.
7. S. Balakrishnan and S. Chopra. Collaborative ranking. In WSDM, 2012.
8. K. Christakopoulou and A. Banerjee. Collaborative ranking with a push at the top.
In WWW, 2015.
9. Q. Yuan, G. Cong, Z. Ma, A. Sun, and N. Magnenat-Thalmann. Time-aware point-
of-interest recommendation. In SIGIR, 2013.