Let's say there's two product types that already established, Search goods and Experience goods(Search goods refer to goods whose quality is easily evaluated before purchase. And Experience goods refer to goods whose quality can only be evaluated(experienced) after purchase. for more detail, please refer to here.)
Based on the previous survey, I have a list of goods classified as Search goods or Experience goods. Between the two types of products, as many researchers pointed, there are different features of online reviews that cover those two types of products.(for example, online reviews for search goods are shorter than reviews for experience goods.) Also, I have large amount of reviews for products of the two types.
Now, I want to classify new products into the product types, based on the features of reviews of those products. In this case, I think using regression with dummy coding will be nice.
dummy coding will be:
Type1 Type2 Experience goods 1 0 Search goods 0 1 New product 0 0
Based on this coding, I can check the significance of Type1 and Type2 variable.(I guess, If Type1 is significant and Type2 is insignificant, the new product can be classified as Search goods, vice versa. Or, If both of the Types are insignificant, the new product cannot be classified in that grouping scheme.)
Main point is, I'm not sure that it is persuasive to conduct this modeling in order to classify the new product. If it isn't please let me know better idea for classification. (But regression will be highly preferred.)