14. アソシエーション分析[Agrawal, SIGMOD1993]
?? 頻出の部分集合を抽出
–? Support(支持度)とConfidence(信頼度)という指標を使用
?? BoWの際の設定
n: BoWのベクトル数 (n=5)
X, Y: BoWの非ゼロのベクトル位置(B1, B2 … B10)の集合
(X U Y): XとYをどちらも含む特徴ベクトル (e.g. “(B4 B6) U B9” )
R. Agrawal+, “Mining Association Rules between Sets of Items in Large Databases”, in SIGMOD1993.
n
countYX
support
).( ∪
=
countX
countYX
confidence
.
).( ∪
=
19. 行動予測の課題
- Action sequence
“Walk” => “Sit” の段階で “Using a PC”を予測
- Time zone (補助的な要素)
Day time
???
Daytime
(Time Zone)
Walking
(Previous Activity)
Sitting
(Current Activity)
???
(Next Activity)
xtimezone
xprevious
xcurrent
θ = “Using a PC”
Given
Not given
Time series