Analysis: Machine Learning-based Personalized Recommender Systems

You can read or write phone number list topic here. you can create new topic about the phone number list.
Post Reply
uamita12
Posts: 1
Joined: Sat May 07, 2022 7:23 am

Analysis: Machine Learning-based Personalized Recommender Systems

Post by uamita12 »

The calculation cost of the recommendation model is high, and the recommendation cost of completely relying on the model is too high. Therefore, it is necessary to design a recall strategy to screen out the content candidate set for recommendation from the massive data. simple recall; User history consumes similar content. History: There are long-term interests and immediate interests of users. Long-term interests refer to the interests that users have expressed on the platform in the past week or the past month and one year. Immediate interests got interested. Similarity: In my last article, we can form a content candidate set from both, through content-based similarity, or based on collaborative filtering. Based on the above two dimensions, we can find a candidate set of content similar to a user's historical content.

(1) Matching based on user portraits Segment users through user portraits to collect the hotspots in each segment, such as: IT industry, 24 years old, male, undergraduate, product Wang, you can collect the content that product Wang likes to watch, and also You can read the content that the portrait of a 24-year-old, male after-sales machine likes to watch, Recommendations based on user portraits have two entities: content and users. There needs to be something that connects the two, the label. The conversion of content into tags is called content characterization, and the user is called user characterization. (2) Sort recall Newest, Hottest, Recent, Newly Visited, Manually Featured. (3) Rules phone number list recall Business strategies such as weather, recent searches, purchases by friends, and past habits during the same period. 2. Sorting link (1) Introduction to the model In essence, machine learning is to select algorithms based on existing data, build models based on algorithms and data, and finally predict the future. Simply put, it summarizes the past and predicts the future.

What is a model? Generally speaking, given the value of the independent variable, the value of the dependent variable can be obtained by calculating the expression. In machine learning, the values ​​of the independent variable and the dependent variable are given. Through machine learning, the expression is obtained, that is, the model . In the field of CV, the model can convert the input of an independent variable, that is, a picture, into a classification. In the field of NLP, the model can convert the input of an independent variable, a piece of speech, into text. There are countless combinations of parameters in the model, and we need to find an optimal set of parameters from them. (2) Construction of the model Determine output Y: What is expected by the model, such as: in the field of recommendation, the output we want to get is how likely the user is to click on the recommended content; in predicting the nature of the tumor, we hope to get whether it is benign or malignant; in NLP feature engineering, the text is processed , our desired output is a piece of text.
Post Reply