9 Matching Annotations
  1. Dec 2020
    1. r 2017). Assum-ing the data supplied by the user are unbiased, collaborativefiltering algorithms may still deliver biased recommendationsbecause they tend to push people toward popular products andavoid items with limited historical data (Fleder and Hosanagar2009)


    2. inputs. Marketplaceobservations have also revealedthat several online retailerssteer users toward more expensive products by placing moreprofitable items at the top of suggested purchase lists, evenwhen they are not of high quality (Angwin and Mattu 2016;Hannak et al. 2014; Mikians et al.


    3. 09). Furthermore, when pro-vided with inferior product recommendations, consumersdisplay reactance toward recommender systems by activelyavoiding these suggestions (Fitzsimons and Lehmann 2004),leading to sustained aversion to algorithms when makingsubsequent choices (Dietvorst, Simmons, and Massey2015)


    1. ords. As a result, thereare key demands of recommending products to con-sumers to improve the shopping efficiency and satisfac-tion.


    1. The recommendation system provides recommendations to each user based on their activities, preferences, and behaviors, which are consistent with the user's personal preferences and assist them in making decisions. The social tagging system allows users to assign personal tags based on their own background knowledge in order to share, discover, and recover resources. In addition to the tagging behavior, there is a large amount of valuable information, which strongly indicates that this information needs to be used to provide personalized services[


    1. The recommendation algorithm based on content is simple and clear [14]. The basic idea is composed of the following three main steps: first, abstract the various types of commodities in the platform, and construct a data structure that describes the commodities or attributes of the project. Then, another important individual in the platform, the user, is abstracted, and the user preferences are used as the target for modeling, so as to design the data structure to express the user preferences [15]. Finally, by judging how well the product fits with the user's washing, different logic of design is used to screen out suitable or expected favorite products for the user. In terms of screening method, the simplest is to select the top K items with the highest similarity for the user through traversal, namely KNN algorithm commonly used in machine learning


    1. ; e-commerce sites have an economic incentiveto use personalization to induce users into spending moremoney

      personalize of algorithm can make people spend more money, which help to improve the benifical of both website and merchant.

  2. Nov 2020
    1. Reinforcement learning is a computational approach to understanding and automatinggoal-directed learning and decision making

      Reinforcement learning means automatically set up a goal ttself and continuously optimize the approach in order to get the maximize reward

    2. Roughly speaking, thevalueof a state isthe total amount of reward an agent can expect to accumulate over the future, startingfrom that state.

      the value of a state means the predictive reward that the agent expecting for