Recommendation system

Dear Prof.Choi:

My research interest is mainly the application and optimization of big data and artificial intelligence technology in recommendation system. Recommendation system is a subclass of information filtering system, which presents items that users may be interested in according to their preferences and behaviors. The recommendation system tries to predict what you like about an item, thus recommending an item that you are likely to like.

In March 1995, Robert mstrong et al. from Carnegie Mellon university proposed the personalized navigation system Web Watcher at the American association for artificial intelligence, and MarkoBalabanovic et al. from Stanford university launched the personalized recommendation system LIRA at the same conference. After more than 20 years of rapid development, it has now realized the personalized recommendation of information and products that users are interested in to users based on their information needs and interests.

Recommendation system mainly has three important modules: User modeling module, recommendation object modeling module and recommendation algorithm module.The recommendation system matches the information of interest demand in the user model with the feature information in the recommendation object model, and at the same time uses the corresponding recommendation algorithm to carry out calculation and screening,find the recommendation object that the user may be interested. And through the following five recommendation methods: (1) content based recommendation (2) collaborative filtering recommendation (3) association rules based recommendation (4) utility based recommendation (5) knowledge based recommendation, to present the items that users may be interested in.

The results of the existing recommendation system have low computational credibility and insufficient interpretability, making it difficult to convince users. The system processing object complexity is high, scalability is not good. When the number of new users increases, the recommendation record will be inaccurate. It brings a lot of inconvenience to our life experience. I plan to add hybrid recommendation methods to the existing recommendation system to make up for the shortcomings of each algorithm. The hybrid approach I plan to use is as follows:

(1)   Weighted fusion: the results calculated by multiple recommendation technologies are Weighted and blended to generate recommendations.

(2)   Switch: different recommended technologies are adopted according to the background of the problem and the actual situation.

(3)   Mix: Mix the results recommended by various recommendation algorithms together, and the difficulty is how to reorder them.

(4)   Feature Combination: combine features from different recommendation data sources and use another recommendation technique.

(5)   Cascade: use the latter recommendation method to optimize the former one.

(6)   Feature Augmentation: the output of the former recommendation method serves as the input of the latter, and it differs from the cascade type in that it produces, instead of direct recommendation results, certain features of the recommendation of the following level.

(7)   Meta-level hybrid: it deeply integrates different recommendation models at the model level, rather than just taking one output as the input of another.

Through the study and research of the above mixed recommendation methods, the existing recommendation system will be made more intelligent and more humanized. At the same time, I will actively learn English, improve their English listening, speaking, reading, writing level, and use the fastest time to adapt to the learning environment. I hope professor Choi can give me a chance to learn. I will actively publish the same paper as my research plan and actively cooperate with team members in the team.

I hope professor Choi can give me an opportunity to study in your laboratory.

Yours Chen Fei