5. Match  and inference of experiment

Near preference attribute ot near preference attribut

Sex Size of clothes      Color Category  of clothes

In experiment, we treated sex and the size as body attribute information as data of clothes ,and we treated the  color  and the  category  of clothes  as preference

Body attribute

Preference attribute

attribute information as data of clothes[3].

Sex  man

Example

Size of clothes   M

Color  gray

Category of clothes  J

Preference attribute

Color

Fig. 7. Inference model; This shows Inference model. The inference

model used is approved from three layers.

The bottom of the layer is an input layer  and personal attribute information is  input  there.  The system takes out two products from data base that is extracted by the body attribute, and infers. The data of the product of the sex, the size, the color, and the category of clothes is stored in inside layer. A product near preference attribute information and a product not near  preference  attribute  information  are  output.  The

the another product and  the  product  near  the preference attribute in the same way.  The system decides the most suitable product for  preference attribute information in the product’s data base that is extracted by the body attribute by repeating this work. Finally, remaining product is presented to the user as a recommendation  product of the shop.

6. Experiment  and result

Fig. 6. Classification; This shows classification of body attribute and

preference attribute, and the category of clothes in this research.

We think the recommendation of the product not suitable for the body attribute even if the preference attribute is suitable is meaningless in experiment. Therefore, first, we have the system infer the product’s data base of the shop based on user's body attribute and, make the product’s data base that is extracted by the body attribute. Then, we have the system infer the product’s data base that is extracted by the body attribute based on the preference attribute. Finally, the most suitable product for personal attribute is decided.

The inference model is approved from three layers.

We experimented on two patterns. In subject 1, user has body attribute information that user's sex is a man, and the size of user's clothes is M, and user has preference attribute information that user is interested in the black, and user is interested in jacket group and long sleeve inner group.

In subject2, user has body attribute information that user's sex is a woman, and the size of user's clothes is S, and user has preference attribute information that user is interested in the white, and user is interested in skirt group and short sleeve inner group.

returned with mail. Even in this case, the user can  see the image of the product user  pointed.

6.3. Evaluation  of experiment

In experiment we questioned eight people about this system. The question is whether liked the recommendation product or not, and about utility of system. We had the testee evaluate the system by five stages.  The  result  like  figure14,15  was  obtained  as a

result.  There  was  a  testee  who  had  evaluated  two in

Fig.   8.  Experimental   subject;   This  shows  body   attribute  and 服务机器人系统英文文献和中文翻译(3):http://www.751com.cn/wenxian/lunwen_80167.html