pith. sign in

arxiv: 1906.12324 · v1 · pith:K7JSQC5Inew · submitted 2019-06-23 · 💻 cs.SI · cs.LG· cs.MM

Cross-Platform Modeling of Users' Behavior on Social Media

Haiqian Gu , Jie Wang , Ziwen Wang , Bojin Zhuang , Wenhao Bian , Fei Su This is my paper

Pith reviewed 2026-05-25 17:19 UTC · model grok-4.3

classification 💻 cs.SI cs.LGcs.MM
keywords cross-platform analysismusic preferenceuser profilingsocial media behaviorBig Five personalityK-means clusteringgenre mood tagsaccount matching
0
0 comments X

The pith

Data from matched user accounts on a music app and a microblog platform shows that music genre and mood preferences correlate with personality traits, gender, location, and interests.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper gathers structured and unstructured data from the same individuals on two platforms to examine how listening habits connect to other personal details. It applies K-means clustering to song tags to form five genre groups and four mood groups, then measures associations with Big Five personality scores and basic facts such as region and user tags. A sympathetic reader would care because the method offers an automatic route to richer user portraits by combining footprints that each platform sees only partially. If the links hold, services could infer traits like location or pet preferences from music data alone.

Core claim

By matching accounts across the music and microblog platforms and clustering music tags, the study finds that music preferences align with real social activities, including stronger folk music interest among users in mountainous regions, stronger pop music interest in urban areas, and greater preference for sad music among dog lovers than cat lovers, while also connecting to Big Five personality dimensions.

What carries the argument

Account matching across platforms followed by K-means clustering on genre and mood tags and subsequent correlation analysis against personality and demographic fields.

If this is right

  • Five genre clusters and four mood clusters can be used to predict specific personality patterns and demographic traits.
  • Location-based differences in music taste can be detected without direct location data by using music listening records alone.
  • Pet ownership preferences can be inferred from mood preferences in music listening histories.
  • The same matching-plus-clustering pipeline can be applied to other pairs of vertical platforms to generate finer user profiles.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Services might combine listening data with social posts to improve recommendation accuracy for content outside music.
  • The approach could extend to predicting offline behaviors, such as travel patterns, from online traces on separate apps.
  • Similar cross-platform linking might help identify biases in single-platform user models by revealing traits that one app misses.

Load-bearing premise

The users whose accounts were linked across the two platforms form an unbiased sample whose music and social data accurately reflect the same individuals' preferences and traits.

What would settle it

Collect self-reported music preferences, personality scores, and location data from a fresh random sample of users and test whether the same genre-mood correlations with region and pet preferences appear at comparable strength.

Figures

Figures reproduced from arXiv: 1906.12324 by Bojin Zhuang, Fei Su, Haiqian Gu, Jie Wang, Wenhao Bian, Ziwen Wang.

Figure 1
Figure 1. Figure 1: Key components of our system. II. RELATED WORK Previous researches suggestthat musical preference is linked to individuals personality and personal identities. Rentfrow et al.[4] concluded that people used preferred music to convey their personalities to others. Little and Zuckerman [5] noted that preferred music style of people could reinforce and reflect their personalities. In addition, Rentfrow et al.,… view at source ↗
Figure 7
Figure 7. Figure 7 [PITH_FULL_IMAGE:figures/full_fig_p004_7.png] view at source ↗
Figure 3
Figure 3. Figure 3: Distribution of five music-genre clusters in all four music-mood groups [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Music preference of genre (left) and mood (right) dependent on gender [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Music preference of genre (left) and mood (right) dependent on age B. Age and Music Preference Based on birthday information of users shared in Sina Weibo, users are devided into two special groups: post-90s and post-80s. Among them, post-90s are born between 1990 and 1999, while post-80s are those born between 1980 and 1989. As shown in Error! Reference source not found., it could be concluded that post-9… view at source ↗
Figure 6
Figure 6. Figure 6: Distribution of music preference of genre among different zodiacs [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Distribution of music preference of mood among different zodiacs. TABLE IV. BSIAC GEOGRAPHIC AND ECONOMIC INTRODUCTION ON CHINA REGIONS OF WEIBO (a) ACS music distribution (b) Rock music distribution (c) Pop music distribution (d) Folk music distribution (e) Balance preference distribution [PITH_FULL_IMAGE:figures/full_fig_p005_7.png] view at source ↗
Figure 10
Figure 10. Figure 10: Proportion of high personality in music genre preference cluster [PITH_FULL_IMAGE:figures/full_fig_p006_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Proportion of high personality in music mood preference clusters. TABLE V. DIFFERENT MUSIC PREFERENCE AND SINA WEIBO TAGS Cluster ACS Rock Pop Folk Balance Hobby Interest Character B. Music Genre Preference and Sina Weibo Tags TABLE V. shows top tags of Weibo of all preference clusters of music genre. As shown in the table, users’ Weibo tags are consistent with preference of music genre. For instance, use… view at source ↗
Figure 12
Figure 12. Figure 12: The difference of music mood preference between cat lovers and dog lovers. VIII.LIMITATIONS The present research is subject to two limitations considering the generalizability of the finding results. One limitation is based on the fact that users on NetEase Music and Sina Weibo are almost young people, thus whether the findings would generalize to other age groups is worth further discussion. Another limi… view at source ↗
read the original abstract

With the booming development and popularity of mobile applications, different verticals accumulate abundant data of user information and social behavior, which are spontaneous, genuine and diversified. However, each platform describes user's portraits in only certain aspect, resulting in difficult combination of those internet footprints together. In our research, we proposed a modeling approach to analyze user's online behavior across different social media platforms. Structured and unstructured data of same users shared by NetEase Music and Sina Weibo have been collected for cross-platform analysis of correlations between music preference and other users' characteristics. Based on music tags of genre and mood, genre cluster of five groups and mood cluster of four groups have been formed by computing their collected song lists with K-means method. Moreover, with the help of user data of Weibo, correlations between music preference (i.e. genre, mood) and Big Five personalities (BFPs) and basic information (e.g. gender, resident region, tags) have been comprehensively studied, building up full-scale user portraits with finer grain. Our findings indicate that people's music preference could be linked with their real social activities. For instance, people living in mountainous areas generally prefer folk music, while those in urban areas like pop music more. Interestingly, dog lovers could love sad music more than cat lovers. Moreover, our proposed cross-platform modeling approach could be adapted to other verticals, providing an online automatic way for profiling users in a more precise and comprehensive way.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper presents a cross-platform analysis linking music preferences from NetEase Music (genre and mood clusters derived via K-means on tags) with user traits from Sina Weibo, including Big Five personalities, demographics, region, and tags. It claims correlations such as mountainous residents preferring folk music, urban users favoring pop, and dog owners liking sad music more than cat owners, while proposing the approach for broader user profiling.

Significance. If the user-matching procedure yields an unbiased sample and the reported associations survive proper statistical controls, the work would demonstrate a practical method for combining vertical-platform data to enrich behavioral portraits, with potential applications in recommendation systems and social-media analytics.

major comments (3)
  1. [Data Collection] Data Collection section: the manuscript supplies no sample size, no description of the username- or activity-based matching procedure used to link NetEase and Weibo accounts, and no account of how missing data or duplicate profiles were handled; without these details the headline correlations (region-genre, pet-mood, BFP) cannot be evaluated for selection bias.
  2. [Results] Results section on cluster analysis: the choice of K=5 genre clusters and K=4 mood clusters is presented without justification (e.g., elbow plots, silhouette scores, or stability checks), and no uncertainty estimates or statistical tests accompany the subsequent cross-platform associations.
  3. [Discussion] Discussion of BFP and tag correlations: the reported links between music clusters and Big-Five traits or self-reported tags rest on an untested assumption that the matched users are representative; no propensity-score adjustment or external demographic benchmarking is described to address possible confounding by platform activity level.
minor comments (2)
  1. [Abstract] Abstract and §3: the phrase 'same users shared by NetEase Music and Sina Weibo' is ambiguous; clarify whether linkage was performed by exact username match, self-reported connections, or other criteria.
  2. [Figures] Figure captions: several figures lack axis labels, sample sizes, or error bars, making it difficult to assess the magnitude and reliability of the reported differences.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which highlight important areas for improving the clarity, transparency, and statistical rigor of the manuscript. We address each major comment below and will revise the paper accordingly.

read point-by-point responses

  1. Referee: [Data Collection] Data Collection section: the manuscript supplies no sample size, no description of the username- or activity-based matching procedure used to link NetEase and Weibo accounts, and no account of how missing data or duplicate profiles were handled; without these details the headline correlations (region-genre, pet-mood, BFP) cannot be evaluated for selection bias.

    Authors: We agree that the original Data Collection section omitted these critical details. In the revised manuscript we will add the total number of matched users, a clear description of the username-based matching procedure between the two platforms, and the methods used to handle missing values and duplicate profiles. These additions will enable readers to assess potential selection bias in the reported correlations. revision: yes

  2. Referee: [Results] Results section on cluster analysis: the choice of K=5 genre clusters and K=4 mood clusters is presented without justification (e.g., elbow plots, silhouette scores, or stability checks), and no uncertainty estimates or statistical tests accompany the subsequent cross-platform associations.

    Authors: The referee is correct that the manuscript did not justify the choice of K or provide statistical support for the associations. We will revise the Results section to include elbow plots and/or silhouette scores justifying K=5 for genres and K=4 for moods, together with appropriate uncertainty estimates and statistical tests (e.g., p-values or confidence intervals) for the cross-platform correlations. revision: yes

  3. Referee: [Discussion] Discussion of BFP and tag correlations: the reported links between music clusters and Big-Five traits or self-reported tags rest on an untested assumption that the matched users are representative; no propensity-score adjustment or external demographic benchmarking is described to address possible confounding by platform activity level.

    Authors: We acknowledge that the manuscript did not explicitly test or adjust for the representativeness of the matched sample. In the revised Discussion we will add an explicit statement of this limitation, include any available basic demographic comparisons between the matched sample and platform-wide statistics, and note the potential for confounding by activity level. We will also clarify the assumptions underlying the reported correlations. revision: yes

Circularity Check

0 steps flagged

No circularity: purely observational correlations from external platform data

full rationale

The paper collects matched user data from NetEase Music and Sina Weibo, applies standard K-means clustering to genre/mood tags derived from song lists, and reports direct empirical correlations with Weibo-provided traits (region, BFP, pet tags). No equations, predictions, or central claims reduce to fitted parameters by construction, self-citations, or imported uniqueness results; all steps are downstream of independent external data collection and standard unsupervised clustering. The derivation chain is self-contained against the collected dataset.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The analysis depends on two chosen cluster counts and the assumption that K-means on tag co-occurrence yields psychologically meaningful groups; no new entities are postulated.

free parameters (2)
  • number of genre clusters = 5
    Set to five for K-means on genre tags
  • number of mood clusters = 4
    Set to four for K-means on mood tags
axioms (1)
  • domain assumption K-means clustering on music tags produces groups that correspond to real user preference differences
    Invoked when the paper forms genre and mood clusters and then correlates cluster membership with external variables

pith-pipeline@v0.9.0 · 5803 in / 1326 out tokens · 24940 ms · 2026-05-25T17:19:47.945751+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

13 extracted references · 13 canonical work pages

  1. [2]

    Sina's Weibo outlook buoys Internet stock gains: China overnight,

    Cao, Belinda. "Sina's Weibo outlook buoys Internet stock gains: China overnight," Bloomberg, Dec. 2012

    work page 2012

  2. [3]

    A room with a cue: personality judgments based on offices and bedro oms[J]

    Gosling S D, Ko S J, Mannarelli T, et al. A room with a cue: personality judgments based on offices and bedro oms[J]. Journal of Personality & Social Psychology, 2002, 82(3):379

    work page 2002

  3. [4]

    Message in a ballad: the role of music preferences in interpersonal perception.[J]

    Rentfrow P J, Gosling S D. Message in a ballad: the role of music preferences in interpersonal perception.[J]. Psychological Science, 2006, 17(3):236-242

    work page 2006

  4. [5]

    Sensation seeking and music preferences[J]

    Litle P, Zuckerman M. Sensation seeking and music preferences[J]. Personality & Individual Differences, 1986, 7(4):575-578

    work page 1986

  5. [6]

    The do re mi's of everyday life: the structure and personality correlates of music preferences[J]

    Rentfrow P J, Gosling S D. The do re mi's of everyday life: the structure and personality correlates of music preferences[J]. Journal of Personality & Social Psychology, 2003, 84(6):1236-56

    work page 2003

  6. [7]

    North, David J

    Adrian C. North, David J. Hargreaves. Music and Adolescent Identity[J]. Music Education Research, 1999, 1(1):75-92

    work page 1999

  7. [8]

    Personality in its natural habitat: manifestations and implicit folk theor ies of personality in daily life[J]

    Mehl M R, Gosling S D, Pennebaker J W. Personality in its natural habitat: manifestations and implicit folk theor ies of personality in daily life[J]. Journal of Personality & Social Psychology, 2006, 90(5):862-77

    work page 2006

  8. [9]

    Predicting personality with social media,

    Golbeck, Jennifer, C. Robles, and K. Turner. "Predicting personality with social media," DBLP, pp. 253-262, 2011

    work page 2011

  9. [10]

    Tag-based user profiling for social media recommendation,

    Hung, Chia Chuan, et al. "Tag-based user profiling for social media recommendation," AAAI Workshop - Technical Report, 2008

    work page 2008

  10. [11]

    Modeling of user portraut through social media

    Gu H., Wang J., et al. “Modeling of user portraut through social media”, in press

    work page

  11. [12]

    A neural network approach to personality prediction based on the big - five model,

    Kalghatgi, Mayuri Pundlik, Manjula Ramannavar, and Nandini S. Sidnal. “A neural network approach to personality prediction based on the big - five model,” International Journal of Innovative Research in Advanced Engineering (IJIRAE), vol. 2, no. 8, pp. 56-63, 2015

    work page 2015

  12. [13]

    Musical Taste Cultur es and Tase Publics.[J]

    Fox, William A.|Wince, Michael H. Musical Taste Cultur es and Tase Publics.[J]. Youth & Society, 1975, 7(2):N/A

    work page 1975

  13. [14]

    Personality Differences of Self -Identified Canine and Feline Lovers

    Sara Braun, Jose Gutierez, Kristen Jolsten, Brianna Olbinski, and Denise Guastello. Personality Differences of Self -Identified Canine and Feline Lovers

    work page