FEMA: FLEXIBLE EVOLUTIONARY MULTI-FACETED ANALYSIS FOR DYNAMIC - - PowerPoint PPT Presentation
FEMA: FLEXIBLE EVOLUTIONARY MULTI-FACETED ANALYSIS FOR DYNAMIC BEHAVIOR PATTERN DISCOVERY Meng Jiang, Tsinghua University, Beijing, China Joint work with Peng Cui, Fei Wang, Xinran Xu, Wenwu Zhu and Shiqiang Yang August 25, 2014 NYC, USA
Meng Jiang, Tsinghua University, Beijing, China Joint work with Peng Cui, Fei Wang, Xinran Xu, Wenwu Zhu and Shiqiang Yang August 25, 2014 – NYC, USA
2
KDD’13 KDD’14
3
4
5
6
time
7
time time
8
time time talk tea break travel sleep Tsinghua WWW’14 Tsinghua KDD’14 Hour Month
9
10
11
author affiliation time
12
time user item t1 t2 t3
13
time user item t1 t2 t3 … user user item item time user item t1 t2 t3
14
time user item t1 t2 t3 time user item t1 t2 t3 … user user item item
15
16
X user item user A(1) A(2) 0~t + user item Δt 0~(t+Δt) λ item
user cluster item cluster user cluster item cluster
X(1) user X(2) item matricizing decompose core tensor projection matrix update
ΔX L(1) L(2) user item item regularize
user
17
X user item user A(1) A(2) 0~t + user item Δt 0~(t+Δt) λ item
user cluster item cluster user cluster item cluster
X(1) user X(2) item matricizing decompose core tensor projection matrix update
ΔX L(1) L(2) user item item regularize
user
18
19
20
21
22
Microsoft Academic Search Tencent Weibo MAE RMSE MAE RMSE FEMA 0.735 0.944 0.894 1.312 EMA 0.794 1.130 0.932 1.556 EA 0.979 1.364 1.120 1.873 Precision vs Recall X L X X Predict “Who”-“What keyword” FEMA uses “Where” (affiliation). Predict “Who”-“@Whom” FEMA use “What” (tweet word).
23
Microsoft Academic Search Tencent Weibo MAE RMSE MAE RMSE FEMA 0.893 1.215 0.954 1.437 EMA 0.909 1.466 0.986 1.698 DTA [Sun et al.] 0.950 1.556 1.105 1.889 Precision vs Recall X L X “Who”-“Where”-“What keyword”? “Who”-“@Whom”-“What”?
24
Evolutionary analysis: update λ and a with ΔX Re-decompose updated matrices Evolutionary analysis: update λ and a with ΔX Re-decompose updated matrices Insensitive to regularization weight
25
26
27
28
29
30
31