DEIM Forum 2009 E5-3 464-8601 1 606-8501 464 8601 1 E-mail: lifushi@arch.itc.nagoya-u.ac.jp, mayumi@mm.media.kyoto-u.ac.jp, {hirano,kajita,mase}@itc.nagoya-u.ac.jp Abstract Study on a Recipe Recommendation Method to Reflect Personal Preference and Nutritional Information Fushi LI, Mayumi UEDA, Yasushi HIRANO, Shoji KAJITA, and Kenji MASE Graduate School of Information Science, Nagoya University Furo-cho 1, Chikusa-ku, Nagoya, Aichi, 464-8601, Japan Graduate School of Informatics, Kyoto University Yoshida Nihonmatsu-cho, Sakyo-ku, Kyoto, Kyoto, 606-8501 Japan Information Technology Center, Nagoya University Furo-cho 1, Chikusa-ku, Nagoya, Aichi, 464 8601 Japan E-mail: lifushi@arch.itc.nagoya-u.ac.jp, mayumi@mm.media.kyoto-u.ac.jp, {hirano,kajita,mase}@itc.nagoya-u.ac.jp There are many web sites recommending cooking recipe. They present cooking recipe in the order of entry date, access frequency or user s ratings, which can not reflect personal preferences and health condition. We are engaged on developing recipe recommendation method based on personal use history of foodstuff to reflect personal preferences. However, we consider nutritional information is important for healthy life. In this paper, we propose a recipe recommendation method which considers dietary preference and nutritional information of foodstuff. Our new recommendation method is re-ranking recipe ranking based on nutritional information. Key words Cooking History Recommendation System, Recipe Recommendation, Nutritional Information, Personal Preference, 1. Web NHK [1] 3 [2]
2008 10 500 2 6000 Web 46 13,391 1 [3] TF-IDF (Term-Frequency - Inverted Document Frequency) FF-IRF (Foodstuff Frequency - Inverted Recipe Frequency) (FF) (irf ) [3] (1) (2) (3) [4] [5] [4] [6] 6 100g [7] 1 2009 1 8 1 2. 3. 4. 2. 2. 1 [3] Web 1 2 F k 3 k 4 Score i 5 6 7 (1) 2. 2 F k F K TF-IDF (Foodstuff Frequency) (Inverted Recipe Frequency) ( (1))
F k = f k irf k (1) k f k k W c = (c 1)/c ( (2)) t W c 1 (k i c), f k = c=1 0 (k i c ). c(1 < = c < = t) i c c W c (1) (2) 1 3 6 7 f (2) 2 f = 7 c=1 c 1 c 1 = 2.36 k irf k ( (3)) irf k = log 10 (M k /M) (3) M M k k M = 465, 325 M = 13, 391 irf = 1.54 F k = f irf k = 3.64 i Score i F k ( (4)) Score i = t c=1 F k irf k irfk (4) 2. 3 FF-IRF [3] 1. 13,391 10 90 20 2 3 2 4 ( (5)) = 25 (1) (2) 2 (3) 3 5 2 3 2 365 [8] 10 2 26 (5)
3. FF-IRF [3] [4] (2005 ) [9] 17 5 3 5 [9] (2005 ) a ) (2005 ) 18 29 I( ) 1,750kcal II( ) 2,050kcal III( ) 2,350kcal I( ) 2,300kcal II( ) 2,650kcal III( ) 3,050kcal 1 3 17 45 5 6 (2005 ) b ) (2005 ) 18 29 20 30 (% ) 4 c ) ( ) (2005 )18 29 40g/ 50g/ 20 (% ) 5 50 97 98 1 1 d ) () (2005 18 29 50 70 (% ) 6 17g/ 21g/ e ) 16 ( ) 4 ( ) (%) = (g) 9/ (kcal) 100 5 (%)= (g) 4/ (kcal) 100 6 (%)= (g) 4/ (kcal) 100
4 5 18 29 600mg( 1.5g ) 8g 700mg 600mg 2300mg f ) 1 A B1 B2 C 18 29 C 85mg 100mg 4. (2005 ) [9] 3. 4. 1 4 2 1 ( 1 ) 2 1 3 2 4 2 ( 2 ) 5 ( 1 ) 2 6 7 3. 1 k 2 k k k j j Score j k k Score j = Score i + Score i (6) Score i Score i k k Score i Score i Score i = Score i = Fk irf k irfk (7) Fk irf k irfk F k irf k 2. 4. 2 (8) 1 ( (5)) 1 2 3 4 5 6 2.
7 8 DB 4. 3 1 ( 1) 1 1 DB 2 B Vol.40(20050331) pp.223-228 2005. [7] Web.IA Vol.103 No.62 pp.1-6 2003.5. [8] 365 2001. [9] (2005 ) http://www.mhlw.go.jp/houdou /2004/11/h1122-2.html 2004.10. 5. 2 [1] NHK http://www.- nhk.or.jp/partner/ryouri/ [2] 3 http://www.ntv.co.jp/3min/ [3] Letters Vol.6 No.4 pp.29-32 2008.3. [4] http://www.mhlw.go.jp/shingi/2004/12/s1202-4.html 2004.3. [5] http://www.mhlw.go.jp/bunya/kenkou/eiyou.html [6] Web
1 (kcal) (g) (g) (g) (mg) (g) (20 ) 583 25 17 95 300 3.3 315 14.1 7.33 45 17 2.6 576 32 28.33 43.5 84 0.3 661 36.6 26.33 62.1 58 2.3 524 28.8 16.33 58.7 85 5.3 422 21.1 11.33 55.2 45 2.6 530 25.3 20.33 58.2 41 1.3 302 12.5 5.33 48.9 29 1.8 708 38.4 16.33 95.6 679 5.8 648 43.8 29.33 44.9 23 1.9 493 26.8 16.33 58.4 280 4.7 2 (kcal) (g) (g) (g) (mg) (g) 315 14.1 7.33 45 17 2.6 287 18.4 18 10.3 223 1.5 602 32.5 25.33 55.3 240 4.1 (20 ) 583 25 17 95 300 3.3