國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
,
, , ‡ ‡ ‡
“ “ “÷ ÷ ÷
1.1 M M M
®W>§≤Ô Social Network åT \s Collaborative Social Network ãÇ0GitHub å Quora Ñ˙˛ ∫⌘ãÀ(>§≤Ô-⌥÷∫q Ê\ H
⌃´|dÑÛ’ ÂX ÂÊ˛ ↵q ÑÓ⇡⇥q ÑÓ⇡⇢8✏Nq Ê\
X2´ Repository „z¯ OLÜÊ˛⇥ Ñ(6 User å H Project KìÑí’g ±+ (6Ñqˇõ⌦o ⇡#Ñ⌦oÔ˝⇤ T
\≤Ô-Ñv÷∫ ⇣u'Ñqˇ⇥⌃êÇdÑ>§qˇõ ÔÂ⌘⌘ q T \Ñ’Kä å ≤Ô-(6Ñqˇõ≥≠Ñπ✏ Ù⇢Ü„[2, 3, 4]⇥
—tÜ>§≤ÔÑv- |˛Ü˝ œ (⇧Ñqˇõ&∫ö⇡. \
≤Ô- ‹uÑ√⌥⇧ÑÕÅ'[7, 8]⇥⌥≥q>§≤Ô Ñ0π (º>§T
\≤ÔÑâ 0(6 Bã|⇢↵ H ⇡õ H˝„hÜ≤Ô-(6Ñq Ó⇡⇥ãÇ GitHub /ã|∫·|d§í&(–õX2´⌦T Â\Ñs⇥
⇢N⇡õX2´ ã|∫·⌥v¢{ÑË/–§f HX2´Ñ¡ ⇧ û À
⇣Ü ↵˙º HÑT \≤Ô⇥
1.2 v v vÓ Ó ÓÑ Ñ Ñ
∫Ü H0UI⇡õ±+º≤ÔÑ(6 HÑqˇõ⌦o (d÷á-–˙Ü
↵π’ ✏N⌥>§T \≤Ô⌥®¶˚q^‘Üœ (6Ñqˇõ⇥⌘
⌘(≥qÑ˙º(6⌥ HÑ®¶˚q ⌥˙º¢{⇧⌥ Hí’‹¬Ñ>§
T \sKì2L^‘⇥(,á-–˙ÑF∂↵ T N˛ Collaborative Filtering ⇡.´„€(º®¶˚qÑÇı (fö ↵>§T \≤Ô d Çı ºœ ≤Ô-(⇧qˇõÑπb WÕÅÑ\(⇥
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
⌘⌘✏N_h⌃„ Factorization Machines; FM [13] ◆Ù!ã⌥,œ í’‹¬Ñ[(qˇõ<⇥y%Ñ0π(º ‡∫ FM -yµÂ↵ Feature Engineering Ñ ,' Generality åH;' Flexibility ì⌘⌘ÔÂ⌥
GitHub >§≤Ôg(⇧Ñ↵✏º⌦o Âyµ⌘œÑπ✏ e FM !ãÑx“
N↵ œ œ↵¢{⇧⌥ HKìÑ‹¬qˇõ⇥
‧
G⇥⌘⌘ç∫⇡/ ↵;LdbÑqˇõ ⇡✏sW(6úa⇡↵;L⇥ Lu Liu et al. 2012 [8] –˙Ü ↵_á ⇣!ã Üœ ;LdbÑqˇõ⇥⇡⇧
vÑÓÑ(º⇣,(6/&⇤úa–↵;L &˝ œ ˙>§≤Ô-–
↵;LÑqˇõ⇥d rv Myers, Zhu and Leskovec 2012 [12] Ñ ÷Ü
‧
Betweenness Centrality ¨Ô+-√' Markov Centrality Âœ ↵‘(œvqˇõBÑPÀl'[10, 17, 19]⇥Ñ v÷v⌫f⌥(6KìÑq \⇧
‧
Factorization Machines; Rendle and Steffen 2012 [13]
ó’Úì⇣∫®¶!ã-ÑALÄS⇥‡∫É ≈ÔÂ!ÏT N˛ÑT\⌦o ˝ ⌥!Ï-Ñ(
!xx<⇠⌅r!T\Ñ‹¬⇥€Âq™ ⇡✏sW(6B ✏NrX2´qˇÜ (6A ⇥⌘⌘ç∫(6⌥⇧ÓѧíÈc-!Ï0óÑ<s/≤Ô-[(Ñqˇ õ⇥
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
,
, , ‡ ‡ ‡
v v vπ π π’ ’ ’
3.1 ® ® ®¶ ¶ ¶˚ ˚ ˚q q q⌥ ⌥ ⌥> > >⇤ ⇤ ⇤T T T \ \ \≤ ≤ ≤Ô Ô ÔÑ Ñ Ñ^ ^ ^‘ ‘ ‘
T N˛ Collaborative filtering /®¶˚q„€°(ÑÄS⇥( ,Ñö©
⌦ T N˛/ ↵…1N˛wÚ ⌅ ˙º⇢↵ HKìÑT \‹¬ |
˛(6⌥ HKì±+‹¬ÑN↵⇥T N˛ÑN↵- ≈ÂS(6 HÑg
π /✏NZ ⇢↵(6Ñx«ÜÍ’⇣,(6 v÷ HÑO}⇥
(T N˛ÑÇı⌦ ,v⌥>§T \≤Ô-Ñ(6⌥ HKìÑ‹
¬^‘⇣®¶˚q⇥w‘Ü™ ⌘⌘⌥>§T \≤Ô-Ñ ¢{⇧ H Ñ‹
¬ …0®¶˚q-Ñ (6 H Ñ‹¬⇥•W⌘⌘✏N_hx“Ê\T N˛
Ñó’ aœœ ¢{⇧ H ‹¬-Ñ[(qˇ< åœ ˙œM¢{⇧
(≤Ô-Ñ>§qˇõ⇥
‧
Contributor 1 Contributor 3 Contributor 4
Project 1 Project 2 Project 3 Project 4 Project 5
a
‧
‧
target score Contributor Project API information
associated with
‧
_h-order-2 factorization machine Ñö©Ç↵@: [13]⇢
ˆ
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
⇡·Ñ k 2 N+0 /ö©_h⌃„N↵-Ñ≠¶ÑÖ√x( hyperparameter )⇥
(˝✏ (3.5)-Ñ0 ˝✏ ˆy(x) ⌘⌘ö©œ↵¢{⇧ ci Ñqˇõ⌃xÇ
↵⇢
S(ci) = Xm
j=1
ˆ
y(x(ci,pj)), (3.7)
(⌃„_h0óå[(qˇõKå ⌘⌘⌥œ↵¢{⇧Ñqˇõ⌃xö©
∫ œ↵ H pj fà¢{⇧ ci Ñ[(qˇõÑ=å⇥
‧
‧
kennethreitz 1 48976 49 71 11604 772
vinta 2 28822 9 9 871 217
jkbrzt 3 28738 13 20 900 878
donnemartin 4 26277 10 13 4098 519
nvbn 5 25739 32 32 3263 162
rg3 6 23164 7 7 1013 406
minimaxir 7 20112 11 11 128 41
josephmisiti 8 20065 21 21 463 196
Valloric 9 13157 4 4 1435 119
fchollet 10 12947 4 5 2900 271
chrissimpkins 11 12860 21 21 2957 20
drduh 12 10294 2 2 186 26
faif 13 10227 5 5 97 57
soimort 14 10036 5 6 1645 79
tomchristie 15 9577 18 31 5602 625
binux 16 8673 6 6 812 36
apenwarr 17 8601 9 9 912 41
jonathanslenders 18 8594 16 18 2844 93
miguelgrinberg 19 8252 39 43 924 119
alexjc 20 7480 4 4 152 3
fxsjy 21 7422 7 7 422 31
p-e-w 22 7260 5 5 107 5
coleifer 23 7162 23 27 4064 175
mitsuhiko 24 7099 61 74 1140 579
nvie 25 6371 18 24 2069 217
DanMcInerney 26 6147 37 37 190 37
eliangcs 27 6127 11 13 825 66
stephenmcd 28 6026 27 30 6347 574
alex 29 4124 41 80 5163 786
toastdriven 30 1789 38 52 779 457
h4.1: Git Awards ≤Ÿgí M30 Ñ Python ã|∫· Æ⌥q «ô
!x'ºÄª<Ñ!D⌥Wˆ e˙!Ñ©«⌦⇥ÈxÑPú⌥ü,254,265 .Ñ!D⌥Wˆ ⌘Û1,075 . &Âd 1,075 .!D⌥WˆÑ Æ ö∫,v Ñ↵✏º©«⌦ çÂ⌘œÑb✏†e˙!ÑN↵⇥
ÅË✏Ñ/ ˙!ÑN↵-Ñ©«⌦ (6⌥ H⇤2L↵%Ñq ⇥(
6©⌦o-Ñyµ< ∫r(6 \NÑ HX2´- (0Ñ&∫1,075 . ÆgÑ!D⌥WˆÑ (!xcè ⇥ HÑ©⌦o-Ñyµ< ∫ r HX2´ (0Ñ&∫ 1,075 . ÆgÑ!D⌥WˆÑ (!xcè
⇥
‧
W v-√xdÜÌ„!x-∫200 v÷√xÜ∫libFM ⇣-Ñÿç<⇥dd K 1º PageRank ⌥ FM ó’Ñ®_' ,áÑÊWx⁄Ü∫ 20 !¯ √ xÑÊWPúÑsG<⇥4.3 U U U0 0 0⌥ ⌥ ⌥⇡ ⇡ ⇡
ÊWPúÑU0(Üi.U0í ѯ‹¬x˝✏ Spearman’s Rho (⇢) [11]
åKendall’s Tau (⌧) [5, 13] ÜU0ÊWPúÑ*£ ⌦i.ÑU0π’-í h ˛ } ⌥ fi Ñ < À º 1 Û -1 Kì⇥ Â-c∫Ñí P ú ∫ X = {x1, x2, . . . , xn} ⇣,Ñí Pú∫ Y = {y1, y2, . . . , yn} Gí PúÑ*£ù
↵⌫i↵˝✏⌃% ó⇢
4,á(PageRank ÑÊW- damping factor Ñ<-ö∫ 0.85⇥
‧
⌧ = #concordant pairs #discordant pairs 0.5· n · (n 1)
U0⌥⇡ H=x H=x PageRank !API⌦o DAPI⌦o Top 10 ⇢ 0.7697 0.8182 0.7818 0.8958 0.9533
⌧ 0.6000 0.6444 0.6000 0.7533 0.8711 Top 20 ⇢ 0.3805 0.4045 0.4601 0.5252 0.3938
⌧ 0.2842 0.3053 0.1684 0.4458 0.3337 Top 30 ⇢ 0.0207 0.0029 0.1604 0.0927 0.0242
⌧ 0.0046 0.0207 0.1172 0.1195 0.0575 h4.2: M30 Ñ⇣,PúU0⌃x
ÊWπ’ FM !API⌦o FM DAPI⌦o
U0⌥⇡ sG< äpx sG< äpx Top 10 ⇢ 0.8958 0.001954 0.9533 0.000283
⌧ 0.7533 0.007584 0.8711 0.002474 Top 20 ⇢ 0.5252 0.001070 0.3938 0.002217
⌧ 0.4458 0.000703 0.3337 0.001692 Top 30 ⇢ 0.0927 0.002697 0.0242 0.000573
⌧ 0.1195 0.000664 0.0575 0.000316 h4.3: 20 !ÊWU0⌃xÑsG<⌥äpx
(ÊW- ⌘⌘⌥⇣, óÑí ⌥Git Awards ≤Ÿ⌦Ñí 2L‘⇤⇥
‧
numpy⇤ 0.167783150 sqlalchemy.types† 0.141090137 twisted.cred.portal§ 0.140377013 pygments.highlight§ 0.133651760 flask.current app§ 0.136406151 decimal.Decimal 0.127058477 django.core.paginator.Paginator§ 0.133800450 asyncio 0.121183140 django.contrib.sitemaps.Sitemap§ 0.128256130 jinja2.runtime.Undefined§ 0.119622888 scipy.optimize⇤ 0.127342390 django.views.generic.View§ 0.116999270 weboob.tools.backend.Module§ 0.122519923 django.test.TestCase§ 0.115921046 curses 0.120499280 sqlalchemy.ForeignKey† 0.110107542 constants.eStart 0.112403762 module base.ModuleBase 0.110074160 core 0.111670590 django.template.loader§ 0.109540859
h4.4: M10 Python modules/packages ц⌦⌦Õ
Ù2 e¿flPython modules/packages Ñ©«⌦( FM !ã◆ÙB†⌦Ñ
⌦ÕPú h4.4-ÑPú ÂOrÜ\∫ (⌦Ñ⌃^
1. —x ó!D⇤ Scientific computing modules 2. «ô´…(¯‹!D† Database-related modules 3. ≤ ⌥≤ŸÑ…(ã|!D§ Web-related modules
v-⇧r⇡⇠Ñ!D⇢´(º—x ó Xr⇡⇠Ñ!D⇢´(º«ô
´ Database Õ\¯‹ Õr⇡⇠Ñ!D⇢´(º≤ ⌥≤ŸÑ…(ã
| h4.4Ñí PúÔ¿fl˙ ≤ ã|¯‹Ñ!D/h-Ñ;Å!D 1 dÔÂÓM⇥ ÑPython ã|∫·@‹ËÑã|Ô˝⇢∫≤Ÿ⌥≤ …(⇥
5ÊWÑPú⌥‘⇤˙ñÑoW'¢W Significance Test Pú P <Ü✏º 0.05
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
4.5 ñ ñ ñ∫ ∫ ∫
4.1: ñ∫ ≤Ÿ∂À
(,‡¿- ⌘⌘–õÜ«ô∆-M10 Python ã|∫·Ñí ⇣,PúÑñ∫
≤Ÿüã6 ⌥«ô⌥ÊWPú¿fi≤ÔÑb✏Zñ∫H˛ &(¿flKåx
˙⌥‡¿4.4¯|…ÑPú ñ∫ Â≤ Ñb✏\H˛ Ç 4.1@: ≤ Ñ(π✏⌃∫ ↵@J⇥
1. (6⌥ HÑ‹¬≤Ô 2. s0«⌦
3. ⇣,π’Ñ⌥€
((6⌥ HÑ‹¬≤Ô- (6Â√rÑ¿fih: HÂÕrÑ¿fih
: qˇõ‹¬Â¿fiKìÑ#⁄h: ⇣,ÑPú (6ÂróÑ⌃xÿ
N ⌃è1ÊÛÛZ*Híè⇥(s0«⌦Ño:@J- o:ÓM⌥⇡x«Ñ
(6 H 1 c∫Ñí Pú⌥ÊWπ’Ñ⇣,Pú⇥,v(ñ∫ Ñ≤
–õÜ . Ñπ’ÜZ⇣,PúÑ‘⇤ .π’⌃%/ Origin FM ⌥ FM †⌦↵✏ºÑ©yµ⇥
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Ç 4.2@: ( Origin π’- o:ÑPú/Â∞ÎNÑ H\∫‘⇤˙
ñ í Ñ⇣,⌃x/q (6∞ÎNÑ H=x (≤Ô-Ô¿fló0Ñ/
¿fiKìÑ‹¬#⁄xœ⌥∆¶ ‡∫íèÑπ’@Ù ⇧Z01ÊÛÛ^ Ñ
˛a _⌥ùd˛aÜ‘⇤v÷i↵π’Ññ∫Hú⇥
(‘⇤ 4.2⌥ 4.3Ññ∫Pú- ÷6π’ FM ⇣,Ñí Pú⇤s F ñ∫⌦Ô¿fl˙#PÑ∆¶Â⇤G˚ ‹ÇÑπ✏H˛⇥ ÕÅÑ/(‘⇤
4.3⌥ 4.4Ññ∫Pú- Ô¿fl˙( M «⌦ÑÊW-ró }Ñí Pú #PÑ⌃H≈b_⌥ 4.2Y∫¯— Â1ÊÛÛ^ Ññ∫Hú⇥
4.2: Â∞ÎNÑ H=xíèÑ⇣,Pú
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
4.3: FM !†eM «⌦Ñ⇣,Pú
4.4: FM †eM «⌦Ñ⇣,Pú
‧
x⁄ q PúÇ h4.5@: ÔÂ|˛v- awesome-python (652070) H Ñ x⌥!DWˆÑ.^xKìÑ‹¬ %ºv÷Ñ H (6vinta ¡ Ñ= x 28,822 - H awesome-python (652070) 1 x28,640 ThËÑ 99.4% ( †eM «⌦ÑÊW- awesome-python (652070) HÕÔ´ñ∫
↵c8Ñ H F/(†eM «⌦ÑÊW- ‡∫awesome-python (652070) HÑ!D⌥Wˆ (x∫0 M «⌦Ñ⌦Õó(6 å†=Ñ⇣,⌃x'E
↵Ó @ÂÊWPú" 'E¶Ñ e≈b2 qˇt‘ÑM 10 ⇣,⇥(d
¿flKå⌥¡ ÿ =x F/!˚U (!D⌥Wˆ / (=xu⌘Ñ≈
b Ç Hawesome-python(652070) ñ∫ ºã|…(ÑPython H ⌥d HÑ'Íx^º«⌦YtÑ^ã⇥
BlackWidow(652070) 64 65 2 5 13
Haul(652070) 54 55 2 80 15
anobii2douban(652070) 4 4 1 3 6
awesome-python(652070) 599 980 571 28,640 0 django-email-confirm-la(652070) 85 101 4 11 64
pangu.py(652070) 36 39 55 4 4
python-pay2go(652070) 8 8 1 3 13
python-smsking(652070) 3 3 1 2 6
sbi.py(652070) 18 19 2 23 8
h4.5: (6 vinta ¡ Ñ HÑq «⌦
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
,
, , î î î ‡ ‡ ‡ P P P÷ ÷ ÷
5.1 P P P÷ ÷ ÷
,v–˙Ü ↵π’ )((6å⇧ÓÑ◆oÜ!Ï>§T \≤a-Ñ qˇõÙc ✏N⌥>§T\≤a⌥®¶˚q^‘Üœ (6Ñqˇõ⇥w‘
Ü™ ⌘⌘⇢N,œ˙º⌃„_h FM ÑF∂↵ œ↵(6 ºœ↵⇧ÓÑ[
(qˇÜœ (6(>§-Ñqˇõ⇥&ÂPython û ∫; ( GitHub ≤Ÿ
-Ñ«ô∆2LÊW ÊWÑPúIÊÜ@–˙Ñπ’Ñ H' ∫⇥ Ñ
Python ã|∫·–õÜ‘~.˙ñπ’Ù}Ñí ⇥ddK ✏N†e©
·o FM !ã2Ù eT⇣Ü 'E¶ÑH˝h˛ s(í M10MÑã|∫
· s ⇢ = 0.9533 ⌥ ⌧ = 0.8711⇥
dÜ⌃êÊWPúÑx⁄K Âñ∫ ≤ ÑH˛Pú &⌃ê†e◆Ù KåÑ!ã-ÑM «⌦ º>§T\≤aÑ flå⌃„/^8 ˘<Ñ⇥
*Üπ⌘⇧φev÷û .^Ñ«ô∆Êã ãÇ ÜÍGitHub Ñ Java «ô
∆ ⇥ddK (*ÜÑÊW-⌥⇧Ïå⌃êv÷ º!D⌥Wˆ©·o
ãÇ –§ÑBì·o / H ÷@ÑáW⌦o ⇥
‧
[1] S. Brin and L. Page. The anatomy of a large-scale hypertextual web search engine. In Proceedings of the 7th International World Wide Web Conference, pages 107–117, 1998.
[2] M. Cha, A. Mislove, and K. P. Gummadi. A measurement-driven analysis of infor-mation propagation in the flickr social network. In Proceedings of the 18th Interna-tional Conference on World Wide Web, pages 721–730. ACM, 2009.
[3] D. Gruhl, R. Guha, D. Liben-Nowell, and A. Tomkins. Information diffusion through blogspace. In Proceedings of the 13th International Conference on World Wide Web, pages 491–501. ACM, 2004.
[4] L. Hong, O. Dan, and B. D. Davison. Predicting popular messages in twitter. In Proceedings of the 20th International Conference Companion on World Wide Web, pages 57–58. ACM, 2011.
[5] M. G. Kendall. A new measure of rank correlation. Biometrika, 30(1/2):81–93, 1938.
[6] N. Li and D. Gillet. Identifying influential scholars in academic social media plat-forms. In Proceedings of the 2013 IEEE/ACM International Conference on Ad-vances in Social Networks Analysis and Mining, pages 608–614. ACM, 2013.
[7] A. Lima, L. Rossi, and M. Musolesi. Coding together at scale: Github as a collabo-rative social network. arXiv preprint arXiv:1407.2535, 2014.
[8] L. Liu, J. Tang, J. Han, and S. Yang. Learning influence from heterogeneous social networks. Data Mining and Knowledge Discovery, 25(3):511–544, 2012.
[9] X. Liu, J. Bollen, M. L. Nelson, and H. Van de Sompel. Co-authorship networks in the digital library research community. Information Processing & Management, 41(6):1462–1480, 2005.
‧
[10] P. Mutschke. Mining networks and central entities in digital libraries. a graph theo-retic approach applied to co-author networks. In International Symposium on Intel-ligent Data Analysis, pages 155–166. Springer, 2003.
[11] J. L. Myers, A. Well, and R. F. Lorch. Research design and statistical analysis.
Routledge, 2010.
[12] S. A. Myers, C. Zhu, and J. Leskovec. Information diffusion and external influence in networks. In Proceedings of the 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 33–41. ACM, 2012.
[13] S. Rendle. Factorization machines with libfm. ACM Transactions on Intelligent Systems and Technology (TIST), 3(3):57, 2012.
[14] X. Shuai, Y. Ding, J. Busemeyer, S. Chen, Y. Sun, and J. Tang. Modeling indirect in-fluence on twitter. International Journal on Semantic Web and Information Systems (IJSWIS), 8(4):20–36, 2012.
[15] L. Terveen and W. Hill. Beyond recommender systems: Helping people help each other. HCI in the New Millennium, 1(2001):487–509, 2001.
[16] M.-F. Tsai, C.-J. Wang, and Z.-L. Lin. Social influencer analysis with factorization machines. In Proceedings of the ACM Web Science Conference, WebSci ’15, pages 50:1–50:2, New York, NY, USA, 2015. ACM.
[17] S. White and P. Smyth. Algorithms for estimating relative importance in networks.
In Proceedings of the ninth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 266–275. ACM, 2003.
[18] E. Yan and Y. Ding. Discovering author impact: A pagerank perspective. Informa-tion Processing & Management, 47(1):125–134, 2011.
[19] L.-c. Yin, H. Kretschmer, R. A. Hanneman, and Z.-y. Liu. Connection and strati-fication in research collaboration: An analysis of the collnet network. Information Processing & Management, 42(6):1599–1613, 2006.