1 (n = 52, 386) DL (n = 52, 386) DL DL [4] Dynamic Time Warping(DTW ) [5] Altmetrics Gunther [

Size: px

Start display at page:

Download "1 (n = 52, 386) DL (n = 52, 386) DL DL [4] Dynamic Time Warping(DTW ) [5] Altmetrics Gunther ["

はるまさいまいだ
5 years ago
Views:

1 DEIM Forum 2014 C {sugiyama-iori@ed., ishikawa-hiroshi@}tmu.ac.jp, endo-masaki@ed.tmu.ac.jp, yokoyama@inf.shizuoka.ac.jp (bibliometrics) h-index Dynamic Time Warping BIRCH 1. [1] 1 (bibliometrics) Twitter Facebook mendeley (Altmetrics) altmetric.com ImpactStory Altmetrics Altmetrics Ceek.jp Altmetrics 1 Altmetrics 1 Accessed 2013 [2] Altmetrics (DL ) DL 3 DL

1 (n = 52, 386) 1.0000 0.0695 1.0000 2 DL (n = 52, 386) DL 1.0000 DL 0.2852 1.0000 1 2. [4] 1 1 12 Dynamic Time Warping(DTW ) [5] Altmetrics Gunther [6] Twitter Twitter [7] Web 3.

2 1 (n = 52, 386) DL (n = 52, 386) DL DL [4] Dynamic Time Warping(DTW ) [5] Altmetrics Gunther [6] Twitter Twitter [7] Web 3. Public Library Of Science(PLOS) 2 PLOS API PLOS Web (html views) 2 Accessed PDF (pdf views) DL DL PLOS (citations) r( n = 52, 386) r = ( 1) 1 DL r( n = 52, 386) r = ( 2) 2 3 DL k-means Dynamic Time Warping(DTW: ) CF

3 3 DL 5 I = J 4 ( ) [8], [9] 4. 1 k-means DL, A A = a 1, a 2,..., a I x DL y 5 x 1 y a k+1 a k A V A (1) (2) V A = { v A(1), v A(2),..., v A(k),..., v A(I 1) } (1) v A(k) = (1, a k+1 a k ) (2) k-means A B v A(k) v B(k) θ k (3) 0 < = cosθ k < = A B V A V B s cos sum cos prod s = (cos sum, cos prod ) cos sum (4) cos prod (5) cosθ k = v A(k) v B(k) v A(k) v B(k) (3) I 1 cos sum = cosθ k (4) k=1 (0 < = cos sum < = I 1) I 1 cos prod = cosθ k (5) k=1 (0 < = cos prod < = 1) s k-means k DOI(Digital Object Identifier) k 1 (6) V m N V m = 1 N V n N n=1 { 1 N = v n(1),..., 1 N N n=1 } N v n(i 1) n=1 (6)

7 ATW 6 DTW 4. 2 4. 2. 1 Dynamic Time Warping Dynamic Time Warping(DTW: ) 1 A B 12 4. 1. 1 1 A = (a 1, a 2,..., a 11, a 12 ), B = (b 1, b 2,.

4 7 ATW 6 DTW Dynamic Time Warping Dynamic Time Warping(DTW: ) 1 A B A = (a 1, a 2,..., a 11, a 12 ), B = (b 1, b 2,..., b 11, b 12 ) A B A B I J A B f k = (a ik, b jk ) A B A B F = (f 1, f 2,..., f k,..., f K ) 6 a ik b jk δ(f k ) (7) δ(f k ) F (F ) (8) w k f k (9) δ(f k ) = a ik b jk (7) (F ) = 1 K w k δ(f k ) I + J (8) k=1 w k = (i k i k 1 ) + (j k j k 1 ) (9) i 0 = j 0 = 0 (F ) A B (F ) F A B (F ) 2 DTW Python mlpy mlpy.dtw std (DTWS ) [8], [9] 2 BIRCH CF DTW CF CF N 0 LS 0 = N 0 k=1 X k 2 SS 0 = N 0 k=1 (X k) 2 CF = (N 0, LS 0, SS 0 ) DTWS DT W S CF 2 SS 0 DT W S = (N, AT W ) N DT W S AT W AT W 2 A B AT W β = (x 1, x 2,..., x k,..., x K ) x k f k (10) AT W β A B (K > = I = J) AT W = (y 1, y 2,..., y k,..., y I ) x k 1 x k 0.5 AT W β AT W y k x k (11) (12) 6 7 x k = ai + k bj k (10) 2 { xk (11) y k = x k + x k+1 2 (12) DTWS CF DTW PLOS 3 Dynamic Time Warping (DTW)-mlpy v3.4.0 documentation, Accessed 2013

5 3 Data1 Data ,261 10, , ,363 10, , ,934 10, , DL ,555 52,386 52, DL Data Data DL Data1 Data2 Data Data1 10% k-means k-means k = 10, 25 DL 3 DL k = 10 4 k = > = 10 (%) k-means k = Kruskal-Wallis Kruskal-Wallis 4 10-means ( ) Data1 Data2 > = 10 (%) > = 10 (%) > = 10 (%) 13,659 2, , ,542 1, , , ,392 1, , , , , , R 4 Kruskal.test Kruskal-Wallis 6 6 p 1, p 2 < = α = DTWS DTWS DL DTWS X DL X = {0.3, 0.5, 0.7, 1, 1.5, 2, 3, 5, 7, 10, 13, 15, 17, 20, 22, 25, 27, 30, 32, 35, 37, 40, 42, 45, 47, 50} DL 26 2, X html DL X pdf 4 Accessed 2013

6 means ( ) Data1 Data2 > = 10 (%) > = 10 (%) > = 10 (%) 4, , ,761 1, , , , , , , means Kruskal-Wallis χ 2 p Data p 1 < ** Data p 2 < ** Clst % > = 90% Num 3 DTWS 3 32 DL 47 Data1 80% 42 DL 15 Data2 72% DL 50 Data1 93% 50 DL 40 Data2 79% 1 F P R F F (13) (14) (15) P R (15) β β = 3.5 R(%) = 100 (13) 7 3 DTWS ( ) Data1 Data2 X html X pdf Clst 10 > = 90% Num > = 90% Num , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,817 P (%) = 100 (14) F (%) = (β2 + 1) P R β 2 P + R (15) F 10 F % 90% 8 F F F 3 F 10 F % 75% 11 F F 3 6

7 8 F (β = 3.5) DL F ,470 Data1 8, ,353 Data1 8, ,447 Data1 8, ,766 Data1 8, ,486 Data1 8, ,276 Data1 8, ,250 Data1 8, ,464 Data1 8, ,524 Data1 8, ,311 Data1 8, F F (β = 3.5) DL F Data Data Data Data Data Data Data Data Data Data F k-means DL Kruskal-Wallis DTWS 90% 80% 1 75% 90% 90% 75% DL 3 Data Data1 3 Data2 8 3 DTWS 8 77% 98% DTWS 3 Data1 Data DL k-means DTWS 2 DTWS 3 k-means k 3 k-means x-means k DTWS DTWS 73% 90% 70% 90%

8 DTWS DL PLOS DL PLOS PLOS 12 DL Altmetrics pp [8] pp [9] pp [1] Open Access (1) SPARC Japan news letter (14) pp [2] pp [3] Altmetrics Vol.55-No.9 pp [4] [5] Hiroaki Sakoe Seibi Chiba, Dynamic Programming Algorithm Optimization for Spoken Word Recognition IEEE Transactions on Acoustics, Speech, and Signal Processing, Vol.ASSP-26-No.1, pp43-49, 1978 [6] Gunther Eysenbach Can Tweets Predict Citations? Metrics of Social Impact Based on Twitter and Correlation with Traditional Metrics of Scientific Impact Jounal of Medical Internet Research, Vol.13-No.4-e123, [7] Twitter Vol.113-No.105

DEIM Forum 2014 B Twitter Twitter Twitter 2006 Twitter 201

DEIM Forum 2014 B Twitter Twitter Twitter 2006 Twitter 201 DEIM Forum 2014 B2-4 305 8550 1 2 305 8550 1 2 E-mail: {yamaguchi,yamahei,satoh}@ce.slis.tsukuba.ac.jp Twitter Twitter 2 1 1. Twitter 2006 Twitter 2012 5 [1]Twitter RT RT Twitter Twitter RT Twitter 2 1