社群協同合作平台之推薦問題研究－以GitHub為例 - 政大學術集成

全文

(1)↵À?ª'x«⌦—x˚ Department of Computer Science National Chengchi University ©Î÷á Master’s Thesis. >§T. \sK®¶OLv Â GitHub ∫ã. A Study of Recommendations on Social Collaboration Platforms: Using GitHub as an Example.  v ⇢ P ⌥ Yà⇢!òZÎ ãÁ9ZÎ. -Ô⌘↵ ~ˆm t ] September 2017.

(2) 106. © Î ÷ á. > § T \ s  K ® ¶ O L  v \ Â GitHub ∫ ã. ? ª ' x « ⌦ — x ˚.  P.

(3) >§T. Â GitHub ∫ã. \sK®¶OLv. A Study of Recommendations on Social Collaboration Platforms: Using GitHub as an Example v. ⇢. P. Student⇢Chia-Yu Tsui. ⌥ Yà⇢!òZÎ ãÁ9ZÎ Advisor⇢Dr. Ming-Feng Tsai, Dr. Chuan-Ju Wang. 立. ↵À?ª'x 治 «政 ⌦—x˚. 大. ©Î÷á. ‧. ‧ 國. 學 sit. y. Nat. A Thesis submitted to Department of Computer Science. n. er. io. National Chengchi University in partial afulfillment of the Requirements iv l C n forhthe i U e ndegree g c h of Master in Computer Science. -Ô⌘↵ ~ˆm t ] September 2017.

(4) Ù. v@Ñit( CLIP ÊW§. Ñx“0à⇢⇥ñHÅ. !ò. +⌥ãÁ9 + v ÊW§µxBfÜ⌘⇡↵_⇤ ì⌘⇡it ˝ (iM +Ñ0√Y ↵⇣w ( CLIP U /⌘Bx ØEÊÑit⇥Ê À. dÜ. +. Ñ1/ÊW§Ñxws◊ ⌥óÚ 治政大 ⇢Ñ∫⇥ Ñ /f⌘( mÂX⌦k©. Åy% ÷⌘U. 立. ‧ 國. ∫. v@it `⌘Ña⌘j4 Ô⌦˛ìk©⌘⌥yU⌘Ñ. 學. ⌥⌘ F S¸ÑÊW§}%4 çõÊ_ÔÂ™õ!N⇥ å. _/‡∫ `⌘⌘M˝ ⇣w Ô⇥. . ‧. n. al. er. io. sit. y. Nat. ↵À?ª'x«⌦—x˚ September 2017. Ch. engchi. 1. i Un. v. P.

(5) \sK®¶OLv Â GitHub ∫ã. >§T. -áXÅ ,÷á–˙ .(>§T\s GitHub ⌦Ñ®¶π’ )(>§ T\s⌦Ñ«⌦º⌃„_h ( Factorizaction Machines !1 FM ) ! ã-⇥ ñH ⌘⌘Ω÷ HÑT\‹¬ HgÑáW⌥↵✏º v \/yµ«⌦†e!ã-◆Ù 2 Â!ãÑ◆ÙPúªZ®¶⇥⌘ ⌘)( GitHub s⌦ ã|⇧ HÑL∫ ( Ç fàÑ _ ‹ Ë ã⌃/Ë/⌥¢{ ) ª˙Àã|⇧ H £U⇢ &" (⇧⌥ HÑ‹¬ÈcÜv\⌘⌘Ñx“Ó⇡⇥ …1⇡#Ñπ’ ⌘⌘ ≈˝ k©!ã6ÇÑ˝–G®¶ÑPú ⇡.✏N xœÑ¯<yµπ’ÑÔk©(⇧•¯0Ù⇢b⌘Ñi¡⇥(ÊW ÷/†eáWyµÑ/↵✏ºyµ ¯⇤º≥qÑ®¶π’T N˛ ⌘⌘(sGæ∫G< ( Mean Average Precision, MAP) Ïfi á ( Recall ) ⌥ F1 ⌃x ( F1 score ) ↵U0↵˝ ⇤*¿Ñh˛⇥ å ÊWPúo: (⇡.T\ã| HÑ GitHub >§T\s⌦ dÜ ,áW«⌦ ↵✏º«⌦(®¶⌦/Ù k©Ñyµ«⌦⇥. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 2. i Un. v.

(6) A Study of Recommendations on Social Collaboration Platforms: Using GitHub as an Example. Abstract This paper proposes a recommendation approach based on Factorization Machines (FM) for GitHub, a social collaborative platform for program development. This work first extracts several features related to collaboration relationship and textual information within the project and the codes, and then incorporates the features into the FM model for training and learning. After the training, the learned models are utilized for recommendation. This work utilizes the behaviors of developers toward a project, such as the star labeling, watch, fork, and contribution, to establish the degree of interest of a developer toward the project. Then, the proposed approach follows the construction of User-Item matrix for conducting the FM learning process. This approach not only expedites the convergence speed and the accuracy of FM, but it also enables users to explore the objects from different aspects. In the experiments, we compare the proposed approach with the traditional collaboration filtering methods in terms of Mean Average Precision (MAP), Recall and F1 measures. The experimental results show that the proposed method outperforms the traditional user-based and item-based collaboration filtering methods. Furthermore, the experiment shows that, for social collaboration platform for program development, the incorporation of code feature is of greater enhancement than textual feature in the task of recommendation.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 3. i Un. v.

(7) Ó⌅ Ù. 1. -áXÅ. 2. Abstract , ‡ À9 . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 M . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 vÓÑ . . . . . . . . . . . . . . . . . . . . . . . ,å‡ ¯‹á{¢ . . . . . . . . . . . . . . . . . . . . . . 2.1 ®¶˚q ( Recommendation Systems ) . . . . . . . 2.1.1 T N˛ ( Collaborative Filtering ) . . . . . 2.1.2 ˙ºgπN˛ ( Content-Based Filtering ) . 2.1.3 ˜ ãó’ . . . . . . . . . . . . . . . . 2.2 ⌃„_h ( Factorization Machines ) . . . . . . . . , ‡ vπ’. . . . . . . . . . . . . . . . . . . . . . . . . 3.1 ,Ñ⌃„_h ( Standard Factorization Machine ) 3.2 >§T\sÑ®¶F∂ . . . . . . . . . . . . . . 3.3 ˙ºgπyµ ( Content-based Feature ) . . . . . . 3.3.1 ,áWÑyµ«⌦ . . . . . . . . . . . . 3.3.2 ↵✏ºÑyµ«⌦ . . . . . . . . . . . . . ,€‡ ÊWPú⌥ ÷ . . . . . . . . . . . . . . . . . . . . 4.1 ÊW-ö . . . . . . . . . . . . . . . . . . . . . . . 4.1.1 «ôê∆ . . . . . . . . . . . . . . . . . . . 4.1.2 ÊW«ô∆ . . . . . . . . . . . . . . . . . 4.1.3 U0⌥⇡ . . . . . . . . . . . . . . . . . . . 4.2 >§T\s®¶˚q . . . . . . . . . . . . . . . 4.2.1 ˙ºT N˛Ñ®¶ . . . . . . . . . . . . 4.2.2 ⌃„_h†eáWyµ . . . . . . . . . . . 4.2.3 ⌃„_h†e↵✏ºyµ . . . . . . . . . 4.2.4 ⌃„_h†eáWyµ⌥↵✏ºyµ . . . 4.3 O ¶,f . . . . . . . . . . . . . . . . . . . . . . 4.3.1 ±⌘œÑ≠¶ k . . . . . . . . . . . . . . . 4.3.2 !ã˜LÙ !x . . . . . . . . . . . . . 4.4 ÊWPúK9ÑÕÅ'¢, . . . . . . . . . . . . 4.5 yµ⌦Õ . . . . . . . . . . . . . . . . . . . . . . . ,î‡ P÷ . . . . . . . . . . . . . . . . . . . . . . . . . . .. 立. 政治大. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. n. al. er. io. sit. Nat. y. ‧. ‧ 國. 學. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. Ch. engchi. 4. i Un. v. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3 1 1 2 3 3 3 4 4 5 7 7 7 9 10 11 13 13 13 14 15 15 16 16 17 17 18 18 21 22 22 24.

(8) Ó⌅ 2.1 T. N˛Çı:✏. . . . . . . . . . . . . . . . . . . . . . . . . . .. 3.1 GitHub s⌦ HÑ £«⌦ . . . . . . . . . . . . . 3.2 (⇧⌥i¡Ñ £U⌃ ⌥⇣-⌦Õ< . . . . . . 3.3 >§sÑ‹¬≤a ⌥(⇧åi¡Ñ‹¬ÈcI€ 3.4 H™ áˆ README.ed . . . . . . . . . . . . . . . 3.5 ↵✏º (˝✏´Gµ . . . . . . . . . . . . . . . . . .. 立. ‧ 國. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. n. er. io. sit. y. Nat. al. Ch. engchi. 5. i Un. v. . . . . . .. ‧. «ôê∆∂À . . . . . . . . . . . . . . Awesome Python Logo . . . . . . . . . . . ±⌘œ≠¶ k ⌥ MAP U0Ñ‹¬Ú⁄ . . ±⌘œ≠¶ k ⌥ Recall U0Ñ‹¬Ú⁄ . ±⌘œ≠¶ k ⌥ F1 score U0Ñ‹¬Ú⁄ !ãÙ !x⌥ .U0π’Ñ‹¬Ú⁄. 學. 4.1 4.2 4.3 4.4 4.5 4.6. 政治大. . . . . . .. 4. . . . . .. . . . . .. . . . . .. . . . . .. . . . . .. . . . . .. 8 9 9 10 11. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. 14 14 19 20 20 21.

(9) hÓ⌅ h 3.1 ↵✏ºyµÑ˝✏´˙˛;áq x⁄ Top 10 . . . . . . . . . . . h 3.2 áWgπ ó TF-IDF ⌦Õ< Top 10 . . . . . . . . . . . . . . . . .. 12 12. h 4.1 t‘ÊWPú . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . h 4.2 T ¢,Pú . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . h 4.3 ˝✏⌦Õ Top 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 18 22 23. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 6. i Un. v.

(10) ,. ‡. À9 1.1. M. 政治大. 立. ‧ 國. ¡Í⇥⇡.✏NT Ñú} [11]. N˛Ñπ’. d∫®¶˚q-. (˙ºG-í¯·˚Ñ(⇧w. ¯. ^<. ∫≥qÑ\’⇥. ‧. Ê. 學. ®¶˚q⇠flÑv—tÜ÷óÜ¯v↵¶Ñ2e $v/( 2006 t Netflix ' NKå⇥ÓMÑπ’'⇢˙º(⇧ìÑ·˚‹¬⌥Ù∫ [2, 12] (Ü–G®¶Ñ. ®W>§≤Ô ( Social Network ) T\så¯‹…( ( ãÇ⇢Github. y. sit. Ñ(⇧åi¡KìÑí’. er. H ( Repository ) ÜÊ˛⇥. io. \. Nat. Quora ) Ñ˙˛ (⇧ãÀ(>§≤ÔT\s⌦⌥v÷(⇧q Ê\ /⌃ ´Û’ ÂÊ˛q ÑÂ\Ó⇡⇥( Github ⌦ ⇡õq ÑÓ⇡⇢8✏Nq Ê. al. ⇢8≥T(⇧. ⇣u'Ñq v i n Ch ˇ [3, 6, 7] ⇥⌘⌘ÑÂ\2 e–˙(⇧" Ñá∆⌥üÀº ( Source code ) U i e h n g c N˛ ( Collaborative Filtering ) å ˜÷yµ &⌥⇡õÃoyµt 0≥qÑT ⇡#Ñ«⌦Ô˝⇤. T\≤Ô-Ñv÷(⇧. n. [(Ñqˇõ«⌦. ˙ºgπ ( Context-based ) Ñ®¶π’-⇥ ¯<'/T. N˛(®¶-Ñ8√Çı⇥fö(⇧úaÑi¡. É⌘˝. ⌃Ñ®¶(⇧v÷úaÑi¡⇥ i¡KìÑ¯<↵¶Ô/✏N⇢.π✏2L ,œ ãÊ⌦ Ñ,œ¯<¶π’/ÔÂí‹Ñ⇥ãÇ º>§T\s . õ(⇧úa^<ÂX⇠flÑi¡. Ñi¡⇥⌘⌘ ˙p⇥d ß6 ûÊ. ¯<¶(. v÷(⇧Gúa^<ã|⇧®<. dbÑ«⌦äÜä⇢. ⌘⌘1ä. Ô˝–˙⇣üÑ. ¯<'Ñ«⌦◊0 a ( ãÇ⇢(⇧ /i¡ ) Ñyµ ( feature ) ↵“¶Ü↵ ⇡✏sW ↵ aÑhy'⇥ Qº⇡↵y' ⌘ÔÂ. ✏Nøt¯<¶«⌦ÑÄªÜ˙À ↵⇢CÑ®¶˚q⇥ Çú¯<¶/9⁄i¡q´ ´ç∫/˙ºT ¯<. ¯. N˛Ñ!ã⇥ Ê. ¡sÑ∫xÜaœ πb. Çú(i¡(yµzì-,œ‹¬. G@ó0Ñ!ã⇢8´1∫˙ºgπÑ!ã⇥˜ 1. G@ó0Ñ!ãÔÂ ⌦Xi.Ñ˜. ✏!.

(11) ã_(á{-2Lv⇥y%/—tÜ˙˛Ü⌃„_h ( Factorization Machines ) [14] \∫˜ ✏®¶ [õÑF∂⇥✏NivÑyµ ⌃„_h˝ !ˇ1 ⇢˙ºT N˛ ˙ºgπN˛Ñó’⇥⌘⌘(á{-øÂ|˛ (˛ ˙º ⌃„_hÑπ’-. i¡. µ⌘œÜ,œ⇥6. (⇧KìÑ¯<'⇢8✏Ni¡å(⇧h:Ñy. Êõ⌦ÔÂÈcÑb✏ró¯<Ñ«⌦⇥ (⌃„_hF∂. ↵ ÔÂ)(i¡-œ↵q ˙˛Ñ!✏ ( pattern ) ÜróÙ⇢«⌦⇥d ‡∫ Èch: ÅM Ñ–÷yµ⌘œÑ¯<'«⌦ @ÂÈch:ÔÂ‘Æhh: ↵i¡Ù⇢Ñ«⌦ v✏N⌃„_h1˝ T⇣⇥. 1.2  v ÓÑ ∫Ü ˙Ü. 政治大. HÑ)(⇡õ>§T\s-(⇧[(ÑÑqˇ«⌦ (÷á-⌘⌘– .F∂ dF∂ÔÂ)(⇡õM Ñ«⌦Ü–G®¶˚qÑHú⇥⌘⌘✏. 立. N◆Ù!ãª⇣,(⇧⌥i¡ìí’[(Ñ‹¬' &˙º_h⌃„Ñπ’⇥. ‧ 國. £Ñ↵¶. ,á–˙ÑF∂↵. T. N˛. /®¶-. œ. œ↵(⇧. 學. ¡Ô˝. 2. ↵⇤„€´(Ñπ’. i. (6. y. ÔÂ9⁄(⇧ A ÑÂÄÑàª“c. ˙À. sit. w‹o'⇥ ãÇ. N˛⇡↵b✏ÃåÑÇı/…1(⇧úa^<Ñi¡. Nat. ˙º(⇧ÑT. ‧. i¡Ñ✏ã /‹¬˝ÔÂ8ef®¶˚q⇥ T N˛Ñi. 8ãÑb✏/˙ º(⇧ ( User-based ) ÑT N˛ [5] ⌥˙ºi¡ ( Item-based ) ÑT N˛ [16]⇥ (⇧ A Ñ. ⌅Üö. er. io. ú} 6å✏N⇡õ«⌦Ü↵~0Ê ↵w ¯<àª“cÑ(⇧ B 6å⌥ (⇧ B àªi¡®¶f(⇧ A ⇥(⇧⌥(⇧ìÑT N˛'⇢(º®¶. al. n. iv n C i¡f(⇧ G˙ºi¡Ñ®¶ÔÂ✏N˙ºi¡ÑT hengchi U ¯<↵¶Ü®¶¯<⇤ÿÑi¡⇥. 2. N˛. ì1. ói¡.

(12) ,å‡ ¯‹á{¢ 政治大 T N˛/ ◊aŒ Ñ!ãK ⇥ T N˛!ãN˛âí ✏© ÕÅÑ«⌦ &›Y↵¯<w 立 ✏©Ñ!✏Ü⇣,(⇧L∫⌥ú}⇥—tÜ _hx“ÑÄS–õÜ [õ ®¶˚qÚì´„€Ñ((Fm-⇥(®¶!ãF∂-. ‧ 國. 學. Ñπ✏Ü˜L®¶ dÜT N˛⌥˙ºgπN˛Ñπ’ Ñ ⌥i.π’P Ñ˜ ✏π’⇥(,¿- ⌘⌘9⁄ Ñ¿fi › ¯‹v2LøÂ⇥. ‧. N˛ ( Collaborative Filtering ) å˙ºgπ. er. io. ≥qÑ®¶π’ÔÂ⌃∫i'^⇢T. sit. y. Nat. 2.1 ® ¶ ˚ q ( Recommendation Systems ). n. a l ) ⇥ 1⇢W ÑFm®¶˚q˝/˙º⇡õπ’ ÑN˛ ( Content-Based Filtering iv n C Ç Youtube Amazon [4, 10] (Ñπ’⇥ ÷Â⌦i.≥ h e n g c(⇡↵✏¿⌘⌘⌃% hi U qÑπ’. ✏N›Y⇡#i.π’Ñ*fiÜí‹ü. Ñ. ≥. ˆ8˙. .˜. . ó’⇥. 2.1.1 T T. N˛ ( Collaborative Filtering ). N˛Ñπ’/6∆'œ. ‹(⇧ÑL∫. ˙º^<Ñ(⇧ª⇣,v÷. (⇧Ô˝_ £Ñ⌦o⇥⇡.π’˙º(⇧Nª £ /O} ªG-* ÜÑ(⇧(^<ãi⌦ÑL∫h˛ [17] ⇥1⇢ó’(º®,®¶˚q-( ⇧⌥i¡KìÑ‹¬⇥ ãÇ fi(º. s⌘⌘. k ↵¯—Ñ0E ( k-nearest neighbor ) ⇥T. Ü„(⇧⌥i¡Ñgπ⌦o. N˛*. ⌘⌘Õ6ÔÂ2L®¶⇥ ã. Ç Û⇥®¶ ‚⌘⌘ Ü„Û⇥Ñgπ ⌘⌘ù6ÔÂ⌘(⇧®¶≠>⇧ Æ⇥Çı/H1>§s⌦ (⇧⌥i¡KìÑú}‹¬˙Àw>§≤Ô ç⌥≤Ô. 1(⇧⌥i¡ÑÈch:. å↵~⌥´®¶(⇧¡ 3. ¯. ú}.

(13) 2.1: T. Ñ(⇧ &⌥d(⇧(´®¶ / 5(⇧ ˚q £Ñ‹¬. N˛Çı:✏. aí Ñú}Ü2L®¶⇥Ç ( 2.1 ) ⇡ - q A B iM(⇧⌥ X Y Z. 政治大 Ë˚q ⌘⌘ÔÂ↵0 A B iM(⇧˝ X Y ˚q £ ⌘⌘˝ ⌥ 立 ⇡iM(⇧ñ∫¡ ^< £Ñ(⇧ &ùT N˛ÑÇı⌥ A £Ñ˚. ‧ 國. ˙ ºg πN ˛ ( Content-Based Filtering ). Nat. y. ‧. 2.1.2. 學. q Z ®¶f B (⇧⇥. ⌘⌘˜÷i¡-. al. (w. £Ñ∞i¡. /⌫. Çı^<. er. N˛⇥F(N↵-. n. T. £Ñi¡ª®¶(⇧_Ô˝. io. fùg(⇧Nª. sit. ˙ºgπN˛;Å/✏Ni¡Ñgπ«⌦Ü2L®¶⇥ ˙ºgπN˛Ó⇡. √⇤˘<Ñyµ ( Ç:áW⌦. i Un. v. o...II ) I€⇣⌘œzìÑyµh:b✏⇥6å ⌘⌘✏N i¡ÑU⌃Ü ˙À˙ ↵ì(⇧" Ñi¡⌫h⇥⇢N⇡õyµ⌘œÑ⌦oª↵~ ¯<y. Ch. engchi. µÑi¡ ⌥⇡#Ñ∞i¡ñ∫(⇧Ô˝. 2.1.3 ˜. ãó’. ‡∫˙ºáWgπN˛Ñπ’/˙º(⇧ (⇧NªÑ(ìW ⇧Ñ§íqˇ⇥Ê P ÜT. £Ñi¡2L®¶⇥. i¡↵%«⌦. É:O⇤nv÷. ‡d:O ºi¡Íœ ®< ¿fi⌥^<(ìW( É:⌘(⇧Ñ↵'U0⇥∫Ü„zÂ⌦⇡õ:fi ✏N. N˛Ñπ’ó0Ü9Ñ. (⌘⌘ÑÊWv-. Ù2. ¡ ( item-item ) ⇢Û/(⇧. ‡. M ˙Ü˜. ãÑó’⇥. e¢v(⇧⌥(⇧ ( user-user ). i¡⌥i. i¡ ( user-item ) Ñ§í‹¬⇥ ✏NT. N˛Ñπ. ✏ ˙Àw(⇧⌥(⇧ì1i¡˙ÀwÜÑ⌥‹‹¬ †⌦˙ºgπN˛Ñ Çı û-˜÷˙ÕÅw qˇõÑyµ†e!ã◆Ùv- ó⌘⌘Ñ®¶ 4.

(14) Pú ÙSäÑh˛⇥ '⇢ —Ñ˜ ãó’Ñv ˝/⌥T N˛⌥˙ºgπN˛Ñπ’P ⇥T N˛ÑÓÑ/N˛â ⇢∫åx⁄êKìÑT\Ñ⌦o !✏ ( patterns ) ⇥ ⌃„_h/ÓM ÄS. ‡∫É. H2Ñπ’K. ≈˝. !ÏT. [14] ⇥ró’Úì⇣∫®¶˚q-;AÑ. N˛Ñπ’. Ù˝. πøÑ⌥©«⌦†e!. ã-x“⇥. 2.2. ⌃„ _ h ( Factorization Machines ). (⌃„_h-. ↵Ü⇠⌅(⇧⌥i¡KìÑ^%. ⇢% ( (⌘⌘Ñπ’. £Ñ↵¶ ) Ñ‹¬Èc⇥ 1º'⇢>§⌦Ñx⁄. ö©∫. è Ùí U 治政 ↵§íÈcÜh:*ÂÑI⇢⇥ ⇢⇥ ∫ÜK ⇡↵OL ⌃„_h" ãÇ 大 (⇧ A }ÜÛ⇥ X Fí立}Û⇥ Y §íÈcÔÂ!Ï@ (⇧å@ i ⌃„_hÔÂ⇠O(⇧ A. F∂. Û⇥ Y. £⇥ 9⁄⇡↵. 學. ‧ 國. ¡KìÑI⇢⇥ 6å. ⌘⌘ÔÂ1(⇧⌥i¡Ñ‹¬ÈcÜh:>§≤a⇥⌘⌘⌥U⇢ñ∫. (⇧åi¡KìÑ‹¬⇥ ãÇ. Çú(⇧ A úa q´ÜÍ(⇧ B ÑÛ⇥. er. ⌥(⇧. )# ) Ü!Ï. '⇢x˙ºÈc‡P⌃„ ( Matrix Factorization ) ÑÄS. io. (⇧L∫⇥˛ . û⌅✏⌅#Ñ«⌦ ( ãÇ⇢Mn. sit. Nat. 1⇢v [9, 13, 18, 19] f. y. ‧. G⌘⌘((⇧⌥i¡Ñ‹¬Èc-⇠⌅⇥ €Âq™ ⇡✏sW(⇧ B ✏Nr Û⇥qˇÜ(⇧ A ⌘⌘ç∫§íÈc!ÏÑ˘</(º[(Ñqˇõ⇥. al. i¡ÑÈcÙc⇣5œb✏ ( [(⇧,i¡,⌦↵á]. É. CD⇣ ) (º. n. iv ⌦↵á®¶⇥,á⌫f✏N(⇧ n C h ⇣ÑáW⌦o⌥(>§T\s⌦Ñí’‹ e n g c h i U ⌃„_h ( Factorization ¬˙!⇥ ∫d ⌘⌘°(˙ºÈc‡P⌃„Ñó’ Machine ) [15] \∫⌘⌘Ñx“F∂ ãÇ. û>§T\s⌦. ÉÔÂûpÍÑÜê-–÷'œÑyµ⇥. i¡(T\s⌦Ñ´ö©⇠flyµå¯‹áWyµ. ⌦o⇥ ⌃„_hÚÍ ã⇥ ãÇ e(º*. /—tÜ⌅.®¶˚Ÿ [9] ↵. Jason et al. [19] vÜ‹º✏NbO. H;w. ˆ-õÑ!. ⌘(⇧®¶i¡ÑOL. &. Ñ‡P!ã⇥ István et al. [13] °(˙ºÈc‡P⌃„Ñπ’. (. Netflix Prize x⁄∆⌦°(!ÆÑU⌃å⇣,⇥ÔÂ|˛ ⌥⌃„_h^<Ñ!ã 8ãÑOL/ Å ˚ŸÕ∞- ⇣,!ã˚Ÿ⇥∫Ü„z⇡↵OL Rendle œ Ü. ↵1∫ libFM [15] Ñ⇢(F∂. É˝. ⇣üÑ!ã⇥Ç [15] @: libFM ÇÏÜ˛. ✏Nyµì⌃„_h!Ï1⇢v÷ Ñπ’. Ç⇡ñÈc⌃„ Pairwise. Interaction Tensor Factorization å SVD ( Singular Value Decomposition ) ++⇥ Liangjie et al. [8] Ó9ÜüÀ!ã Â B⇢πbU⌃x⁄∆⇥ ¯⇤K↵ (⇡⇧Â\ -. ⌘⌘Ó⇡/⌥¯<'«⌦ue libFM 5. ⇤. F∂2LÕ'ÓÜ9›Y.

(15) libFM Ñ*fi⇥. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 6. i Un. v.

(16) ,. ‡. vπ’ ⌘⌘–˙Ñπ’. 2. ÓÑ⇥Â↵⌘⌘2. et. 政治大 ,Ñ⌃„_h ( Factorization Machine ). ¯<⌦o⌥F∂ÜUI¯<Ñ!✏ ( pattern ) ÜT0. 立. eÑœ. ⌘⌘π’å( FM Ñ®¶F∂ W2. º. eÑ™ ⇥. ‧ 國. 學. , Ñ ⌃ „ _ h ( Standard Factorization Machine ). ‧. Nat. ⌃„_h ( FM ) ÔÂ✏N⌅.^ÑyµÜT0 Ñ§í⌦Õ. io. al. n. !Ïœ. y. ,Ü™Ñ⌃„_hÔö©∫: u X. u X u X. i i=1 j=i+1 n C i=1 U hengchi. yˆ(x) = w0 + v-ˆ y ∫Ó⇡˝x. x“. sit. yµKì@. ,⌃„!ãÑHú⇥É˝. er. 3.1. &. w i xi +. w0 ∫h@OÓ ( global bias ). v. (3.1). wîj xi xj ,. wi /yµ< xi Ñ⌦Õ. yµ<KìÑ§í\(⇥§í\(Ñ√x wîj Ô´⌃„⇣⇣. G wîj. Ñ§í√. x wîj =< vi , vj >=. k X. (3.2). vif vjf .. f =1. §í¯√xG/∫iù±⌘œ vi ⌥ vj ÑgM. ±⌘œ≠¶ k zö!ãÑ⌥‹↵. ¶⇥ ⌃„_h ( FM ) –˙Ù}ÑF∂(®¶OL⌦ ⌥ ,Èc⌃„ Ñ/ ÉÔÂÙπ◆Ñ†e Ñyµ<&–G®¶Hú⇥ ⌃„_h ( FM ) Ñ0¿À√ ⇤ [15] ⇥. 3.2 T. >§T\sÑ®¶F∂ N˛/®¶˚q8´°(ÑÄS⇥(⌘⌘ÑÂ\7. ›. T. N˛Ñ*‚⌥.

(17) ˙º⌃„!ãP. ⇥1º⌃„_hw. û(⇧å⇧Ó-–÷˙⌦o Le0yµ⌘œ-⇥. o}Ñ8eyµ˙!F∂. ⇡.π✏w. T. ⌘⌘ÔÂÙ•. N˛^<ÑÇı. ÔÂπ◆Ñ. ⌘⌘–˙(>§T\s ( Github ) ✏Ni¡ ( repository ) Ñ⌦oª˙À (⇧⌥i¡KìÑ‹¬Èc⇥&⌥⇡↵‹¬Ècv\!ã◆ÙBÑÓ⇡<⇥ ( Github Ñ>§T\s⌦ ( 3.3 ) ⇧+Â↵€⇧:. ⌘⌘x«Ü!Ï(⇧. £U⇢Ñi¡Ñ⌦o. Ç. 政治大. 立. ‧. ‧ 國. 學. n. Ch. • Star⇢(⇧ÔÂ i¡fà. engchi. £«⌦. er. io. al. HÑ. sit. y. Nat 3.1: GitHub s⌦. i Un. v. ⇥ãÇ>§s⌦Ñfiö⇥. • Watch⇢(⇧ÔÂ i¡fi ‹Ë⇥ãÇ>§s⌦Ñ˝dºá⇥ • Fork⇢(⇧ÔÂ i¡ã⌃/0ÍÒÑzì↵2LÓ9 • Contribute⇢(⇧ÔÂ çÔ⇣∫¢{⇧⇥. ⌘⌘ç∫(⇧dÜfi Star. Ë/⇥. i¡⌦≥ÍÒÓ9ÑGµ↵✏º. Fork. / Watch ⇡õ⇤˝. &✏Nü\⇧. h:(⇧. i¡. £Â (>§s⌦ÑÑT\‹¬_˝ h:(⇧ i¡ ( repository ) –.↵¶ ⌦Ñ £ (⌘⌘ÑF∂-M Ñ⇤n0⇡.(⇧✏Ni¡ÑT\ ‹¬ÜhT–.↵¶⌦Ñú}⇥⌥⇡#Ñ«⌦†e⌘⌘Ñ!ã ¶ Âk©!ãÑx“⇥ ⌘⌘⌥(⇧ F (i, j) =. ⇥ si,j +. û†«⌦ÑPÃ. £ÑÑ↵¶ö©Ç↵⇢. ⇥ wi,j + ⇥ fi,j + ✏ ⇥ ci,j 8. (3.3).

(18) v F (i, j) ∫(⇧ i. j i¡Ñ. £↵¶. s, w, f, c ⌃%∫ Star, Watch, Fork,. Contribute Ñ⇤ó< ( i¡ j í dL∫ ) ⌦Õx< Üøt⌅⇧⌦o7¶⇥Ç ( 3.5 ). , , , ✏ Â⌦€⇧↵%Ñ. Ê. User. Star (0, 1). Watch (0, 1). Fork (0, 1). 3.2: (⇧⌥i¡Ñ. £U⌃. Contribute (0, 1). ⌥⇣-⌦Õ<. 政治大 vu ∫, i ↵(⇧ r. )(Â⌦>§s⌦Ñ⌦o. ∫,j↵i¡ ⌘⌘˝ 立 ⌥⌘⌘ÑÓ⇡‹¬ÈcgπÑö©Ç↵⇢ ⌘⌘✏N>§T\s⌦o˙À‹¬ ç1 ( l✏3.3 ) ⌥‹¬. ‧ 國. 3.3 (a) ). j. I€⇣‹¬Èc (. 學. Ç(. i. ¬Èc⌥∫⌘⌘Ñx“Ó⇡Èc⇥. Ch. Repository #4. y engchi. r1. r2. r3. r4. r5. er. n. User3. F (i, j). Repository #3. sit. Repository #2. al. ⇡‹. ‧. io. User2. Nat. User1. Repository #1. 3.3 (b) ). 1. 0. 1. 0. 0. 2. 0. 3. 0. 2. 0. u3. 1. 0. 0. 1. 2. u4. 0. 1. 0. 4. 0. u1. ui n U. v. User4 Repository #5 (a). (b). 3.3: >§sÑ‹¬≤a ⌥(⇧åi¡Ñ‹¬ÈcI€. 3.3. ˙ º g π y µ ( Content-based Feature ). ( libFM Ñ!ãF∂↵ Í› T N˛ÑHú⌥y' Ù˝ Ñ((⌃„_ hF∂↵o}8eyµÑy'⇥ Â↵i↵✏¿⌘⌘⌃%À9(⌘⌘ÑÂ\-i. Ñ˙ºgπ ( Content-based ) yµ. ⌃% ↵✏ºyµ⌥ ,ÑáWyµ⇥ 9.

(19) ,áWÑyµ«⌦. 政治大 H™. áˆ README.ed. ‧ ⇣áWyµ†e!ã-x“. io. F(>§T\s⌦ Github ÑáW⌦o¯⇤í. al. ⌘⌘ç∫/. sit. 3.4 ) ⇥ ⌥⇡õ1(⇧. ⇣ÑáW⌦o/)(À9i¡ü˝ÑáWî. y. (⇧. Nat. (⌘⌘Ñ«ô∆-. 3.4:. 學. ‧ 國. 立. Ç(. k©Ñ⇥. ‡’⌥èƒÑ⇥ (üÀ«ô⌘⌘. er. 3.3.1. v. n. ≈ H⌥ õ‹⌦ì1MU⌃ªd Çcèh:’ ( Regular Expression ) ªd \(W ( Stopword ) ⌥W9Ñü ( Stemming ) Iπ’⇥ç✏N^; ( Term Frequency,. Ch. engchi. i Un. TF ) ⌥⌃⌘áˆ;á ( Inverse Document Frequency, IDF ) \Æ ^Ñ⌃x ö©∫ tf (t, d) ⇥ idf (t, d) tf (t, d) =. ó TF-IDF ⌦ÕÜv. Ç↵⇢. f (w, d) , max{f (w, d : w 2 d}. idf (t, d) = log. (3.4). |D| , | {d 2 D : t 2 d} |. D ∫«ô∆g=á,∆⇥TF-IDF ⌦Õ/. (3.5). .Kº«ô¢" ( Information Retrieval. ) åáW¢ÿ ( Text Mining ) -8(Ñ†⌦ÄS⇥É/. .q. π’. (ºU0. Æ W^ ºá∆ ↵ûô´- Æ á,ÕÅ↵¶⇥W^ÑÕÅ'⇤®WÉ ˙˛Ñ!x⇣c‘ûw F B⇤®WÉ(á∆-˙˛Ñ;áÕ‘↵M⇥ TF-IDF ⌦Õ8´⌧↵. ŒK( \∫á,⌥(⇧Kì¯‹↵¶ÑU⇢⇥. (⌘⌘ÑÂ\-. ⌥✏N TF-IDF ⌦Õ⌃xª. (á∆gW^Ñ TF-IDF ⌦Õ⌃x. ósG 10. á∆-ÑW^Zíè. &). v\Èxyµ<ÑÄª⇥ ⌥ TF-IDF.

(20) ⌦ÕÿºÄª<Â⌦ÑÜv\⌘⌘ÑáWyµ. ⌥⇡õyµ⌘œ†e!ãx“Ü. –G®¶Pú⇥. ↵✏ºÑyµ«⌦. 立. 政治大. ‧. ‧ 國. 學. n. al. er. io. sit. Nat. 3.5: ↵✏º (˝✏´Gµ. y. 3.3.2. i Un. v. (⌘⌘Ñv- –˙Üà £Ñπ’˙Àyµªœ⌘⌘Ñ(⇧⌥i¡⇥( >§T\s Github ⌦ ;Å/⌃´(⇧ÍÒÑ∞Î /Ê\ÑüÀº ⌘⌘. Ch. engchi. ç∫⇡õ1(⇧⌃´˙ÜÑüÀº Ëo. –.↵¶⌦_/. ì1⇡õ↵✏û ¡Ñyµ⇥. Ñgπ. ÷6. õ⇡ñÑ<✏⌥P6 ª¿fl(⇧Ñ“c. (⇡⌘⌘ª¿fl˙(⇧(∞Î↵✏û ã⇥ Python / Î. ↵¡. /∫^Ñù⇢û. 7'˝✏´Ñ↵✏û. ¯⇤w. F1º≈. ,áWÙ∫¥9⇥. …dªvZœ(⇧. ⌦Ñ“c. ˝ /i. (⇡·⌘⌘Â Python ∫. (∞ÎB˝. ✏N˝✏´Ñ|Î. Ñ„zOL⇥ (⌘⌘ÑÂ\- ⌘⌘)( Python ⇡#ÿ;á(˝✏´Ñy. ' ˜÷œ↵ H-(NÑ˝✏´Ñ;á Â Bag of work Ñπ✏˙À↵✏º Ñyµ⌘œ Z∫k©⌘⌘!ãx“Ñyµ«⌦⇥ ⌘⌘. ó«ô∆g@. ˝✏´v⌘⌘Ñ↵✏ºyµ. ˝✏´sG˙˛;áv\ÈxÄª. ⌘xÿºÄªÑ. ç✏N⇡#Ñyµ⌘œªh:(⇧. å⌥⇡.↵✏ºÑgπyµ†e!ã-x“Ü–G!ãx“ÑHú⇥. 11. /i¡.

(21) Library. ;á. pytest re datetime time collections json unittest logging bcbio.pipeline functools. 414 404 368 328 317 310 281 257 219 208. h 3.1: ↵✏ºyµÑ˝✏´˙˛;áq x⁄ Top 10 tf-idf 治weight 政 ycm 448.309 大. Text. io. y. sit. zappa vim schema path aw classdocutil http codeblock. ‧. Nat. 316.834 297.436 243.761 240.300 228.494 214.995 213.378 197.448 189.652. 學. ‧ 國. 立 ndash. n. al. er. h 3.2: áWgπ ó TF-IDF ⌦Õ< Top 10. Ch. engchi. 12. i Un. v.

(22) ,€‡ ÊWPú⌥. 4.1.1. «ôê∆. 政治大. 立. ‧ 國. ÊW-ö. 學. 4.1. ÷. ‧. ⌘⌘ÑÊW/û>§T\s - Github1 ⌦6∆Ñ˛Êx⁄«ô⇥ Github / N git ⌥‰2LH,ß6Ñfl‘üÀº„° Ÿs⇥Github ⌥ÍÒ∞ÎÑ. º↵✏ã|⇧/↵. H✏N2Xh ( repository ) ⌥üÀ. y. ã|⇧˝. Nat. 3ü˝sÑs. ↵✏. sit. al. HÑb✏√⌥ã|. n. ã|⇧Ñ. &ÔÂ✏NT\:✏. Ch. ã| &˝d ∞Ñ©;H,⇥. engchi. ÊWÑ«ô∆/Hì1 Git Awards2. er. io. º⌃´(s⌦ ìv÷ã|⇧√⇤ /T\ã| å⇣ H⇥Ê s⌦ Ñ–õÜ ˚⌫ (Ñü˝ ãÇfö fl® U÷⇥ã|⇧ÔÂ✏N⌥;v÷. v. Ü¢ñ⇢⌘ã|⇧√⌥Ü. i Un. Git Awards /. ↵✏N Github ã|⇧í. Ñ≤Ÿ⇥/ùgã|⇧ HÑ Star x✏N ( l✏4.1 ) óU⌃ ç⌥ã|⇧ù 0@ /û ⌃Ä%^ågU⌃íè ⇡· S(u) h:ã|⇧ u ( Git Awards ⌦Ñ U⌃ S(u) =. n X. si + (1.0. i. si ∫ u ã|⇧, i ↵ í. ªd⇠‘DT. HÑ Star x. 1.0 ) n. (4.1). G i = 1, 2, 3...., n⇥ ⌘⌘⌥≤Ÿ⌦Ñ Python. ~˙ Github ⌦ Python (. Lí. Ñ Top 30 ã|⇧ ( ,á-. –0Ñ Top 30 Ü ⇧+⇠‘DT ∫h⇤í ) &⌥⇡õã|⇧v\⌘⌘, «ôÑ.P ( seed ) ✏N Github API v33 ªì÷id(⇧⌥ HÑ«ô⇥, d∫ Git Awards ⌦ Python ( Lí. Ñ Top 30 ã|⇧⌥ã|⇧ÍÒ¡. 1. http://github.com http://git-awards.com 3 https://developer.github.com/v3/ 2. 13. Ñ H.

(23) Repository owner (1st floor). User. Repository #1. Contributor (2nd floor). Repository #2. Repository #3. …… ……. Repository #n. User. Repository #1. Repository #3. Repository #2. ……. …… ……. Repository #n. 4.1: «ôê∆∂À. çû⇡õ Ñ«ô∆g. H-Ñ¢{⇧⌥¢{⇧Ñ ,. 4.1 )⇥ (⌘⌘. 1000 H çû⇡ 1000 ↵ H↵Ñ 治政 H⇥ t↵«ô∆ q 大 10841 ↵ã|⇧å 55325 ↵ã 立. d Top 30 Ñã|⇧q¡. 10841 ¢{⇧ª,ådÑ | H⇥. ‧ 國. 學. 4.1.2. Ç(. Hv⌘⌘Ñ,åd. ÊW«ô∆. ‧. y. áÑèö⇥∫ÜM. õ. ≈Å⇢⇠Ñ. sit. HÑ«ô&í. Nat. ⌘⌘∫Ü6∆0⇡õ(1 Python Top 30 ã|⇧ \‹¬ó¢{⇧ (M ✏¿ ™ Ü⌘⌘6∆«ôÑπ’⇥F1º Github s⌦ ºX>⌥⌃´(2XhH. ⌘⌘√g1. n. al. er. io. Github ⌦ã|⇧T\Ë/Ñ Awesome Python 4 ⇥⌘⌘⌥ Awesome Python -. ´⌃. v. ´( Github s⌦2Xh-Ñ HÈxâ &ûÈxåÑ Hª‘ X(⌘⌘ ê∆Ñ«ô- H⌥T\ã|⇧⇥(⌘⌘Ñ6∆«ô- qµÀÜ 143 ↵ HÂ 1972 Mã|⇧ ¬Èc. Ch. engchi. i Un. ⌘⌘⌥✏N⇡õ 1972 Mã|⇧⌥ 143 ↵. –õ!ã◆Ù⇥ (⇡ 1972 ã|⇧-. µÀ⌘⌘6∆«ôÑ.P ( Python. Top 30 developer ) ⌥ 143 ↵X>vM Python ±Ä˝✏´Ñ ⇡#6∆«ô⌥ÈxÑπ’ ⌘⌘›YÜ(>§s⌦Ã •N˛â ’. õ⇢⇠Ñ. (⌘⌘ÑÊW-. H⇥(⌘⌘Â\ÑÕÅ§í‹¬ ì. H⇥(↵↵✏¿⌘⌘⌥À9~↵8´((®¶⌦ÑU0π ⌘⌥✏N⇡õU0π’I. ⌘⌘Ñπ’/. ˛⇥ 4. Hª˙À⌘⌘Ñ‹. https://awesome-python.com. 4.2: Awesome Python Logo. 14. ⇤˙rÑh.

(24) 4.1.3. U0⌥⇡. ⌘⌘°(Ü. .‘⇤´8ˇÜU0®¶PúÑπ’⇢M k ↵sGæ∫G< ( Mean. Average Precisioni, MAP ). Ïfiá ( Recall ) å F1 ⌃x ( F1 score ) ⇥. œ↵ã|. ⇧ ‰ P (k) ∫M k ↵æ∫¶ ( Precision @k ) : Pk p=1 P (k) ⇥ ruo(p) AP (u, o) = , I(u) h:i¡ i ( o ⌫h-´íè(, p ↵Mn. £ ( / = 1, & = 0 ) ⇥ M k ↵sGæ∫G<Ñl✏Ç↵⇢ PU AP (u, o) M AP @k = u=1 , U. H i /&. v- U h:ã|⇧Ñ∆ Ñh˛⇥. 政治大 h:(®¶æ¶⌦. ⇥ k ↵sGæ∫G<äÿ. 立. Ïfiá/(U0®¶˙ÜÑ. H-. clº´ã|⇧. ‧ 國. £Ñ |. H∆. ⌥®¶Ñ. Nat. H∆. §∆xœ. Ù}. /úaÑx. ⇡✏s⇧®¶˙ÜÑ/ã|⇧. (4.4). ⌃ÕG/(⇧. Êúa. io. sit. Hx⇥ ÿÏfiá. H⇥. (4.3). £Ñ. er. £Ñ. H | \ | ®¶Ñ H | , £Ñ H |. ‧. |. Recall = ⌃P∫(⇧. £. º. 學. œ⇥ ól✏Ç↵⇢. £Ñ. rui h:ã|⇧ u. y. v o(p) = i. (4.2). al. n. iv ( Binary C/hF1 ⌃x /(qU n⌃ê-å26⌃^ Classification ) Ñæ∫¶,œ⇥É✏Næ∫¶åÏfiáÜ óåó0 ÔÂ™ F ¶ engchi F ¶œ. ⌘⌘»1 F ⌃x. œ/æ∫¶åÏfiáÑsG< l✏Ç↵⇢ F1 = 2 ⇥  1 => F1 >= 0⇥1 ∫ }. P recision ⇥ Recall , P recision + Recall. (4.5). G 0 ∫ Ó⇥. (ÊW- ⌘⌘‘⇤≥qÑ®¶π’⌥⌘⌘®¶F∂↵ÑÓp çì1Â⌦ –0Ñ .(®¶⌦8ãÑU0⇡ñ ªI (⌃„_hÑF∂↵†egπyµ å. oÑ–ÿ®¶Pú. ∫˙r⇥. 4.2 > § T \ s  ®¶ ˚ q (,¿-. ⌘⌘⌥Õfi>(À9ÇU✏N>§s⌦Ñã|⇧ÑL∫«⌦ª◆Ù. !ã⇥(ÊW-⌘⌘⌥û. Ñ“¶⌥π’Ü–G®¶ÑPú⇥ 15.

(25) 4.2.1. ˙ºT. (⌘⌘ÊWv-. N˛Ñ®¶ ↵U0ÑÕfi. /⌥⌘⌘(>§s⌦Ñ«ô. Ñπ’Z®¶⇥⌘⌘⌥⌃„_h⌥Â↵i.≥qT (⇧ ( User-based ) ÑT. ✏NT. N˛. N˛Ñπ’2L‘⇤⇢ ˙º. N˛å˙ºi¡ ( Item-based ) ÑT. N˛⇥Â↵œ. ⇡õπ’Ñ;ÅÇıÀÛ⇥ • User-based⇢⇡↵π’/⌥(>§s⌦ ⌦. @. &↵~¯<L∫Ñ§‘⌥ã|⇧Ñ∆. ã|⇧|d¯<'2L† ( ã|⇧Kìw. ÿ¯<¶ ). ⇥ É√g¯<Ñ§‘ /ã|⇧ª2L⇣, U⌃⇥‰ R(u) /1ã|⇧ u £ HÑ∆ ⇥ã|⇧ u åã|⇧ v Ñ¯<¶ óÇ↵⇢ R(u) \ R(v) | R(i) |↵ | R(j) |1. , 政治大. su,v =. 立. v ↵ 2 [0, 1] ∫øt√x⇥. (4.6). ↵. ‧ 國. óÇ↵⇢. io. (4.7). y. £Ñ∆ ⇥. n. al. º Hi. ↵,. sit. Nat. v- U (i) ∫Íõã|⇧. U (u) \ U (v) | U (i) |↵ | U (j) |1. ‧. su,v =. er. H¯<¶. 學. • Item-based⇢⇡↵π’⌥˙º(⇧ÑT N˛^< F a ⇥É/ ó HKìÑ¯<' &9⁄ã|⇧Nªú} £Ñ H2LU⌃⇥. i Un. v. 9⁄ MSD (Million Songs Dataset) ⌘0 [1] -ÑPú ⌘⌘⌥˙(⇧ÑT N˛å˙ºi¡ÑT N˛-ö ↵ = 0.5 ⇥ h-⌫˙ÜM k ↵sGæ∫¶ Ïfi áå F ¶œ. Ch. eÑPú⇥Çh@:. N˛Ñπ’<N˝. ⌃. engchi. ⌘⌘ÔÂ↵0dÜ®_Ñ˙ñK. @. T. M k ↵sGæ∫¶˝=( 0.25 0 0.45 Kì⇥û-⌘⌘. ÔÂ|˛ ÷6˙ºi¡ÑT N˛(M k ↵sGæ∫ h˛ ∫Å˙ ⇢Ûÿº ⌘⌘ÑF∂˙ ⌃„_h F⌃„_h(Ïfiá⌥ F ¶œÕ ⇤˙rÑh˛⇥( ⌘⌘ÑÂ\-x«⌃„_h∫⌘⌘ÑF∂˙ Ùπ◆Ñ†eM. /QºÉ¯⇤w≥qÑπ’ÔÂ. Ñyµ<«⌦–G®¶Hú⇥(•↵Ü✏¿. _h⇡↵y'ª7 ⌘⌘Ñ!ã. ⌘⌘⌥✏N⌃„. &ó⌘⌘ÑPú ↵¶⌦Ñ–G⇥. 4.2.2 ⌃ „_ h† eá W yµ (,á- ( 3.3.1¿ ) –0Ü⌘⌘(˙ºgπÑáWyµ (⇧Ñ™. áˆ ( README.rd ) -ÑáW«⌦Üv\. ⌘«ô∆-@. HÑ(⇧™. áˆ˙ÀwáW" 16. /ê∆ã|⇧. H–õ. HÑáWyµ⇥ ñH⌘ ç✏N". ª˙Àœ↵.

(26) HÑáWyµ⌘œ⇥ T\‹¬ Ù⇥h-. Ü. HÑáWyµ⌘œ. ⌘⌘1˝. ùgœ↵ã|⇧Ñ. ⌥ H⌘œZ†=∫ã|⇧ÑáWyµ⌘œ &†e⌘⌘Ñ!ã-◆ ÔÂ↵0†eáWÑyµ ( s FM + text ) ⇥ Ç ( h 4.1 ) ÊWPú⌘. ⌘ÔÂ↵0 i¡ÑT. †eáWyµå. (. .U0π’↵˝. oÑ–G. N˛⇥ 1dÔã†eM. ÑáWyµ©. *º˙º. /. ©º–G!ã◆Ù⌥. (,á- ( 3.3.2¿ ) –0Ü⌘⌘(˙ºgπÑ↵✏ºyµ. ÂÄ⇤(áWÜ. ®¶Pú⇥. 4.2.3 ⌃ „_ h† e ↵✏ ºy µ. v\gπyµ. /°(ã|⇧(∞Î. ✏´(Ñ“c⌥;á\∫. HN↵-Ñ↵✏º. ⌅(⇧. Hyµ⌘œ⇥ ñH⌘⌘✏Nœ↵. H-. 政治大. º˝. ã|⇧∞. Î↵✏º⇠⌅↵ÜÑ˝✏´(ìW˙À HÑyµ⌘œ çùgœ↵ã|⇧T \NÑ H⌘œZ†=∫ã|⇧Ñyµ⌘œ &†e⌘⌘Ñ!ã◆Ù-⇥ û ( h. 立. 4.1 ) ⌦bÑÊWPú⌘⌘ÔÂ↵0. ⇢⌦'EÜ–GÜ 0.16. N˛ 0.13 ↵ÆM⇥(⇡Ë⌃ÊW¶˝ *ºv÷π’Ñ*ph˛ Ê ˛. ⌥⇡.1↵✏º. ⇢ÛÿNÜ˙ºi¡ÑT. ÷/M k ↵sGæ∫¶ Ïfiá⌥ F œ ¯⇤wMbÑáWyµÊW⌘⌘ÔÂ|. ‧. ‧ 國. (M k ↵sGæ∫¶. ⌅↵ÜÑ(⇧ÑL∫“c†e!ã-Ñπ’. (. º!. ⌃„_h†eáWyµ⌥↵✏ºyµ. n. al. Ch. engchi. er. io. sit. y. Nat. ã◆ÙÑU0Pú⌦ ÜÙ}Ñ–G⇥. 4.2.4. FM + code. 學. )å. †e˙ºgπÑ↵✏ºyµ ( s. i Un. v. 1⌦bi↵✏¿ †e↵✏ºyµ⌥áWyµ oÜk©Ü!ãÑ◆Ù⌥x“⇥ ÷/( MAP Reacll / F1 score ÑU0⌦ ¯⇤w ,Ñ FM ⌘⌘˝ ö↵¶⌦Ñ–G›˙(≥q⌦Ñ~.π’⇥(⇡↵✏¿ Ñyµ⌘œ. ⌘⌘⌥⇡i.. †e!ã-◆Ù⇥ ⌘⌘⌫fÑ⌥↵✏ºyµ⌥áWyµ. k© †e. !ã- PúÇ( h 4.1 )⇥⌘⌘ÔÂ↵0 B†eáWyµ⌥↵✏ºyµÑπ’ ( s FM + text + code ) Pú& ∫ Å˙ /Àº⇡i.yµπ’Kì⇥ûÊ WPú. ⌘⌘¿fl0(!ã◆Ù⌦. B⌥i.⌘œ. u†eB. ↵✏ºyµÑ⌘œ. oÑ*ºáWyµ. Ù. óU0ÑPúÀºi⇧Kì⇥(⇡#ÑPú-. ⌘⌘ÔÂ|˛û↵✏º˙Àyµ⌘œÑπ’ (⌘⌘ÑÊW-/. k©Ñ⇥. 1º↵✏º Å⇢NËohM˝ ˜L @Â(∞Î⌦⇤ èƒ P6⇥Ê Õ✓/ ,ÑáW ( GitHub Ñ README î-/⇤í P6Ñ (⇧⇢Û ÔÂ)(. õ⇡fi&_j˝. ó(áWyµÑË⌃¡. H. ⇡#Ñ≈¡⇤–GóáWMU⌃Ñ⌥‹¶. ⇤⇢Ñ‹⌦⇥ ¯⇤w. N⇤¥9Ñ↵✏ºáW«⌦ª. ,áW«⌦. ⌅(⇧L∫/Ù 17. (Ñ. . ⌘⌘ç∫œ⇡#✏. (ÊWPú_I. Ü.

(27) Model. MAP@10. Recall. F1 score. 0.1828 0.2662 0.4532 0.4250 0.5381 0.5867 0.5488. 0.0500 0.1171 0.1053 0.1443 0.1794 0.1967 0.1854. 0.0578 0.1407 0.1260 0.1737 0.2138 0.2344 0.2210. Randomize User-based CF Item-based CF FM FM + text FM + code FM + code + text. h 4.1: t‘ÊWPú º!ã◆Ù/¯v k©Ñ⇥. 4.3. 政治大. ¶,f. 立. ⌘⌘(ÊWN↵-. ∫Ü˝. ì⌘⌘π’. tÑ√xZÜ. õO. Ù}Ñh˛. ¶⌦Ñ,f⇥ Â↵⌘⌘⌃⇣i↵✏¿. c⌃„BÑ±⌘œ≠¶ k ( 4.3.1 ) ⌥!ã˜LÙ. x ( 4.3.2 ). ⌘⌘⌥¿fl. er. io. al. ⌘œ≠¶√x k. n. iv qˇ↵¶⇥ û À-öÑ 2 = C 8 ⌘⌘Â 2 Ñ!π∫n ↵ÆM h 4, n8 ,16 i U g c,32h,64 √x∫2n vn = 1, 2, 3, ..., 6 ( k = 2, e ) ⇥ ÊWÑ ↵✏ºyµ áWyµ⌥i.. ÑU0Pú w. @–G. 4.3 ) ÷6. F1⇣wE¶Ü¿fl. Ñ. ⌘⌘. a⌃%. ,Ñ. B†eÑπ’ ( Merge ) ⇥. ÔÂ↵0øt√x k Ñ⇣w ÷/. !ã◆ÙÑ qÊWÑm.. 2. ñH MAP ÑË⌃ (. È. sit. ±⌘œÑ≠¶ k. (⇡↵✏¿⌘⌘¿fl FM (ZÈc⌃„Bh:Æ. FM. FM (. y. Nat. 4.3.1. ⌃%. ‧. √xÊW↵U0⌃xÑ⇣w⌥ä’⇥. libFM ÔÂø. ›. 學. ‧ 國. Ê. O. ,Ñ FM. /. oÑk©(⌘⌘. †eyµ⌘œÑπ’˝. oÔÂ¿fl0¯⇤w. ,Ñ FM. ⇣. †eyµ. ⌘œÑ⇣wE¶⇤ÿ⇥Ê (⌘⌘ †eyµ⌘œÑπ’- û MAP (√x k ÑO ¶,f- ⌘⌘|˛⌥√x k -ö∫ 32 ⌥ 64 (U0Pú⌦÷6 ⇣w F/Óp&. '. Ü!ãÑ⌥‹¶. ⇣wE¶. ÇMbÑ-ö⇥Ê. (⌘⌘ÑÊW-œ↵ÆMKì@. Âû MAP ÑË⌃⌘⌘ÔÂ↵˙⌥√x k -ö( 32 ˝/ }Ñx«⇥ çÜ Recall ( 4.4 ) ⌥ F1 score ( Ú⁄¯⇤wÜ π’-. ⇡i⇧. ⇤¯. /. W. –0√x k zö. ÅÑBì˝⇣. x⇣w. @. ÷/(Bì B⌦ /Pú 4.5 ) ÑË⌃⌥Mb MAP Ñ. ÑÚ⁄⇥ (⇡Ë˝ÊW⌘⌘. ÔÂ¿fl0√x k û 2 0 32 ÑÆM˝. ^8¯<. ( 3.1 ¿. †eyµ⌘œÑ. å⇣wÑÚ⁄⌥ MAP Ë⌃. ÑÓp/(√x k = 32 ÂåãÀ↵—⇥(†eyµ⌘œÑË⌃ 18.

(28) Recall F1 score ⌥ MAP. W^<Ñ⇣wÚ⁄. ,Ñ FM -1. F(. W. oÑ. ⇥⌘⌘ÔÂ↵0( ,Ñ FM - û√x k = 4 Âå1ãÀ⇣eH˛^ Ñ Ú⁄⇥ ÷6Ú⁄ ®W√x k Ñû† ⌘Ü^ ÑE¶ Fû⇡Ë⌃ÑÊW⌘ ⌘ÔÂ¿fl0( Recall ⌥ F1 score ⇡i↵U0π’⌦ µ⌘œÑπ’/⇤ , FM ÜóÙ. û†√x k. º. †ey. k©Ñ⇥. (√x k ÑO ¶ÊW- ⌘⌘ÔÂ|˛( †eyµ⌘œÑπ’ ºû† √x k ÑÆM¯⇤w ,Ñ FM /Ù k©Ñ⇥ Ê .U0π’⌘⌘Ô oÑ ®W√x k Ñû†˝. 1√x k ÑO. ¶,fÑPú. Ó!ã◆ÙÑ. 'ÑH ⇥. –G†eyµ⌘œπ’ÑPú. 立. $v/( k = 32 ⇥. ⌘⌘⌥⌘⌘ÑÊW√x˝-ö∫ k = 32. 政治大. ‧. ‧ 國. 學. io. sit. y. Nat. n. al. er. |˛. Ch. engchi. i Un. v. 4.3: ±⌘œ≠¶ k ⌥ MAP U0Ñ‹¬Ú⁄. 19. ⌥˝|.

(29) 立. 政治大. ‧ 國. 學. 4.4: ±⌘œ≠¶ k ⌥ Recall U0Ñ‹¬Ú⁄. ‧. n. er. io. sit. y. Nat. al. Ch. engchi. i Un. v. 4.5: ±⌘œ≠¶ k ⌥ F1 score U0Ñ‹¬Ú⁄. 20.

(30) 4.3.2 (M. !ã˜LÙ ↵✏¿⌘⌘›. ∫ 32 ˝ró ö∫ 32. !x √x k ZÜ~↵ÊW. 'ÑH ⇥⌘⌘ˆå⌦. ûÊW-⌘⌘¿fl0⌥√x k -ö. ↵✏¿ÑÊWPú. &⌥Õfi>(!ãÊW-ÑÙ. ÑÙ. !x(U. 0Pú⌦Ñä ⇥Ù -öû 1 25 50 75 100 ... 300 q ÑÊW-ö⇥ (⇡↵✏¿⌘⌘› Ù !x⌥U0PúÑ‹¬Ú⁄ (. ]↵ 4.6 ) 2L. ¿fl⇥. G-⌃%. º]↵. ÑÙ. MAP. !xÑÊW⇥. 6ÇÚ⁄. ÷/Í.U0π’. Ù Ñ✏. (Ù. ºU0PúÜó⇤ O ⇥ 1Ù xøÿ F1¯⇤Ù x 1 ⌥ 50 KìÑ/ oÑ– x-ö( 100 /. sÑ®‚⇥ ⌘⌘ç∫⌥Ù. H Ñ√x⇥. å⌘⌘✏NO. ö∫ 32 åÙ x-ö⇥. x-ö∫ 100 /. 立. ‧. ‧ 國. 學 y. sit. io. n. 4.6: !ãÙ. &. (⌘⌘ÑÊW-⌥√x k 治政 & H  ˝ |Ó⌘⌘Ñ!ã[õÑ D√ 大. ¶,fÑPúóÂ. Nat. al. ⌃%. xÑÊW-⌘⌘Ô. er. G 100 å1. Ü¿fl(. Recall ⌥ F1 score (√x k = 32 Ñ-ö↵. Â oÑ¿fl0¯⇤√x k ó⌘⌘U0PúÑ⇣w/¯ G. !x. ⌥⌘⌘Ñ√x k -. Ch. engchi. i Un. v. !x⌥ .U0π’Ñ‹¬Ú⁄. 21.

(31) 4.4. ÊWPúK9ÑÕÅ'¢, T ¢,88´(Ü‘⇤iD«ô/&. (v⌦. oWÑÓp⇥T ¢,. »1x. t ¢, /1 William Sealy Gosset º 1908 t@–˙(q x⌦ÑG-¢W⇥ ( ⌘⌘ÑÊW- ∫ÜQö⌘⌘Ñπ’œ⇤º‘⇤Ñπ’ oWÑ} (⇡·⌥⌘ ⌘ÑPú(. ⇧U0⇡ñ⌦↵%ZÜ T ¢,⇥. ⇡·⌘⌘‘⇤Ñ. a∫. CF Recall å F1 score ˝/ aí. Óp 6å. ⇧U0-. MAP@10 ∫ Item-based. ÿÑ⌃x. ,ÑFM⇥ ñH˙À[!G-⌘⌘ÑÊWPú⌥‘⇤. ó t <⇥l✏Ç↵⇢ ¯ u0 X t= q. (4.8). S. 政 P治 x 大 ¯= X n 1 n i. 立. i. ‧ 國. 學. (4.9) n ¯ u0 ∫⌘⌘‘⇤ asG XG/⌘⌘ÊWsG n ∫ÊW!x S ∫⇡ñÓ⇥ ó˙ t <çì1Âhó˙ p < ( p  0.05 h:i⇧ÓpoW⇥h 4.4 - ÔÂ↵ 0(⌘Ñœ⇧ÊW˝1T0 p  0.05 oWÓp. ‧. a⇥h 4.2 -. (⌘⌘ÑÊWPúoW*º⌘⌘‘h⇤. U0Pú. ⌘⌘ÔÂóÂ Z ⇤ ⇡⇠∫. p  0.05* MAP@10 Recall@10 0.5381* 0.1794* 0.5867* 0.1967* 0.5488* 0.1854*. al. F1 score 0.2138* 0.2344* 0.2210*. er. io. Features. sit. y. Nat. p  0.05⇥. ‡dû T ¢,-. n. FM + text FM + code FM + code + text. Ch. engchi. i Un. v. h 4.2: T ¢,Pú. 4.5. yµ⌦Õ. (⇡↵‡¿⌘⌘⌥ø(O ¶,fÑ s√x -MPú }Ñyµ⌘œ Library Feature B ⌥( FM !ã◆ÙN↵ºœ.˝✏ýµx“Ñ⌦Õ<2X↵ Ü. &. 20 !ÊW-. ˝✏ýµÑ⌦Õ<†=å÷ Top 10 Ñ˝✏ýµ. ⌃⇣ Contributors ⌥ Project i. Top 10 ⇥û-⌘⌘ÔÂ¿fl0 ✏´. (⌘⌘Ñx“N↵∫Ê¡ û˝✏ýµ⌦Õ⌘⌘↵0 ÷/ Contributor. q. õ‘⇤±ÄÑ˝. ⇤ÿÑ⌦Õ<⇥ õ. £Ñ0π. / Project ÑË˝ 22. ñH⌘⌘ÔÂ¿fl0 ( h4.3 ). ˝ÔÂ↵01 Python ∞ÎÑã>üÀ.

(32) º…(F∂ Django ⇥ ã|≤. ÑF∂ Django. ;Å/ìã|⌥‹Ñ«ôÉ’. ≤ äó!Æ É( Python -/^8nM´(Üã|≤ ÑF∂K ⇥Ê ÑÔ Â↵0 œ/(Ü∞Î≤Ô,Ú↵✏Ñ˝✏´ Pyspider ⇥ ºÓM«ô—xÑÏ √|U («ô⌃ê˝✏´. ˝. «ôê∆/^8ÕÅÑ ∞. ◆Ñå⇣⌅.˚Ÿ. _‡∫ Python w åôÑ. («ô⌃ê-ì8´(Ü∞Î,Ú↵✏ê∆. «ô /v\s,Z õ«ôMU⌃⇥ ®W«ô—xÑjn ¡ PÃ˝✏´Ñ Python _/⇣∫«ô—x∂≈ôÑÄ˝ $v/Ò¶x“Ñãê˝✏´⇡J⇥ Ò ¶x“. É/. Keras ˝✏´. .˙º^ì≤Ôˆ8Ñπ’. (⌘⌘Ñ Contributors Top 10 -Ñ. 1/ Python (^ì≤Ô⌦Ñ±ÄÑãê˝✏´K. x“ Ñ 1 Google Brain ⇠äã|Ñ TensorFlow Ò¶x“-±»´ ÷⌥(Ñãê˝✏´⇥ å. (⇡¿ÑÊW-ÔÂ¿fl0. ⇥Ê. ⇡õãê˝✏´˝/ÓM(. õ‘⇤±ÄÑ˝✏´. œ/ Python Ñ≤. 政治大 x“N↵∫Ê¡ ⇤ÿÑ⌦Õ<⇥Ê _ÔÂû⇡#π✏ÙØö 立 «ô∆Ñ!ã◆ÙN↵- w qˇõ⌥ÿ⌦Õ<Ñ˝✏ýµ. ã|F∂ Django Â ˛(⇤±ÄÑ_hx“⌥Ò¶xÑÑ˝✏´. £ã|⇧ÑÔ˝'⇥. ‧ 國. 學. G H®¶f Ê. ‧. keras.datasets django.db.models phonenumbers BIFFRecords bcbio.pipeline.config utils Contributors numbers pynamodb.models falcon httpie.input django.utils.timezone oauthlib.oauth1.rfc5849 signal playhouse.tests.base ycmd.utils vcr.request Project pyspider.database.base.taskdb select django.utils django.contrib.contenttypes.models mock. n. er. io. sit. y. Nat. al. Ch. engchi. i Un. h 4.3: ˝✏⌦Õ Top 10. 23. (Ò¶. v. (⌘⌘Ñ. (⌘⌘ Python ÔÂc∫Ñ–.

(33) ,î‡ P÷ 政治大. (,÷á- ⌘⌘–˙Ü . FM (®¶⌦ÑF∂ › T N˛Ñ*‚⌥˙º ⌃„!ãÑP ⇥✏N⌃„_hw o}Ñ8eyµ˙!F∂ ⌥⌘⌘–˙1↵. 立. H^%Ñπ’. ⌘⌥®¶OL(>§T\s⌦2L. †e⌘⌘Ñ!ã-◆Ù⇥Ê. 學. ‧ 國. ✏ºgπÜUIã|⇧“c⌥. )(>§T\s⌦ã|⇧✏N. ‧. Ñ \‹¬ å H(s⌦U⇢Ñ¯‹«⌦ª˙Àã|⇧ \⌘⌘Ñx“Ó⇡ ⇢N FM Ñ ó ⌘⌘ó0ã|⇧[( )(⇡↵Ècª2L®¶⇥ FM i(º⇢.yµ. H˙À. £Ñ↵¶ &v £Ñ‹¬Èc. ⇡_✏sW⌘⌘ÔÂûÙ⇢Ñ. ⌘⌘. ✏NáWyµ. y. k©Ñ«⌦⇥(⌘⌘ÑÊW-. Nat. b⌘r÷Ù. ⌘. ì1↵✏º. sit. al. er. io. gπUI«⌦Ñyµ⌥P Â⌦i.yµÑπ’⇥Â⌦ .π’⌥≥qÑπ’‘ ⇤ Púh Ü†eyµ⌘œÑπ’/ )º–G®¶'˝⇥(⌘⌘ÊW- 1. e!ã◆Ù /˝. n. v ÷/( MAP Recall i n Ch / F1 score ˝ ⇤*¿Ñh˛⇥$v/( MAP ÑU0⌦ Ù‘ Baseline }Ñ i U e h n c g Item-based CF ÿ˙Ü 0.1335⇥ 1dÔã ⌥⌘⌘–˙1↵✏ºUI«⌦Ñyµ† ↵✏ºUI«⌦Ñyµ. ∫˙r. ¯⇤wÂ≥qπ’. HÑ–G®¶Pú⇥. (>§T\s⌦. ã|⇧ÑT\‹¬/. ®¶^8. N>§T\s⌦ã|⇧ÑT\‹¬⌥↵✏ºÑ«⌦ ⌦Ñh˛. F⌘⌘Õ!’⌥i.pÍÑyµ⌘œ. k©Ñ⇥÷6⌘⌘✏ HÑ–GÜ⌘⌘(®¶. †e!ã-k©◆Ù⇥}‘. ⌘⌘Ñ↵✏ºyµ†⌦áWyµÑπ’⇥_1/™ ⌘⌘ *Üì↵✏ºÑy µ«⌦⌥ ,ÑáW«⌦˝ ¯¯⇣ u ˙Ù}Ñ®¶Pú⇥. 24.

(34) √⇤á{ [1] F. Aiolli. A preliminary study on a recommender system for the million songs dataset challenge. In Proceedings of the 2013 International Institute of Refrigeration (IIR), pages 73–83. Citeseer, 2013.. 政治大 tic web. In Proceedings of the 2007 International Joint Conference on Artificial 立 Intelligence (IJCAI), volume 7, pages 2677–2682, 2007.. [2] P. Bedi, H. Kaur, and S. Marwaha. Trust based recommender system for seman-. ‧ 國. 學. [3] M. Cha, A. Mislove, and K. P. Gummadi. A measurement-driven analysis of information propagation in the flickr social network. In Proceedings of the 18th Interna-. ‧. tional Conference on World Wide Web, pages 721–730. ACM, 2009.. y. Nat. [4] J. Davidson, B. Liebald, J. Liu, P. Nandy, T. Van Vleet, U. Gargi, S. Gupta, Y. He,. sit. M. Lambert, B. Livingston, et al. The youtube video recommendation system. In. er. io. Proceedings of the 4th ACM Conference on Recommender Systems, pages 293–296. ACM, 2010.. al. n. iv n C D. Goldberg, D. Nichols, B. M.hOki, i UUsing collaborative filtering to e nandg D. c hTerry.. [5]. weave an information tapestry. Communications of the ACM, 35(12):61–70, 1992. [6] 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. [7] 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. [8] L. Hong, A. S. Doumith, and B. D. Davison. Co-factorization machines: modeling user interests and predicting individual decisions in twitter. In Proceedings of the 6th ACM International Conference on Web Search and Data Mining, pages 557– 566. ACM, 2013.. 25.

(35) [9] M. Jiang, P. Cui, R. Liu, Q. Yang, F. Wang, W. Zhu, and S. Yang. Social contextual recommendation. In Proceedings of the 21st ACM International Conference on Information and Knowledge Management, pages 45–54. ACM, 2012. [10] G. Linden, B. Smith, and J. York. Amazon. com recommendations: Item-to-item collaborative filtering. IEEE Internet Computing, 7(1):76–80, 2003. [11] H. Ma, I. King, and M. R. Lyu. Learning to recommend with social trust ensemble. In Proceedings of the 32nd International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 203–210. ACM, 2009. [12] P. Massa and P. Avesani. Trust-aware recommender systems. In Proceedings of the 2007 ACM Conference on Recommender Systems, pages 17–24. ACM, 2007.. 政治大. [13] I. Pilászy and D. Tikk. Recommending new movies: even a few ratings are more valuable than metadata. In Proceedings of the 3rd ACM Conference on Recommender Systems, pages 93–100. ACM, 2009.. 立. ‧ 國. 學. [14] S. Rendle. Factorization machines. In Processing of the 10th IEEE International Conference on Data Mining (ICDM), pages 995–1000. IEEE, 2010.. ‧. [15] S. Rendle. Factorization machines with libfm. Proceedings of the 2012 ACM Trans-. y. Nat. actions on Intelligent Systems and Technology (TIST), 3(3):57, 2012.. sit. [16] B. Sarwar, G. Karypis, J. Konstan, and J. Riedl. Item-based collaborative filtering. n. al. on World Wide Web, pages 285–295. ACM, 2001.. Ch. engchi. er. io. recommendation algorithms. In Proceedings of the 10th International Conference. i Un. v. [17] J. Schafer, D. Frankowski, J. Herlocker, and S. Sen. Collaborative filtering recommender systems. The adaptive web, pages 291–324, 2007. [18] Y. Shi, A. Karatzoglou, L. Baltrunas, M. Larson, A. Hanjalic, and N. Oliver. Tfmap: optimizing map for top-n context-aware recommendation. In Proceedings of the 35th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 155–164. ACM, 2012. [19] J. Weston, C. Wang, R. Weiss, and A. Berenzweig. Latent collaborative retrieval. arXiv preprint arXiv:1206.4603, 2012.. 26.

(36)