WARNING - This is a toy project for course. The tech or tools used here might out of date!! But it is fun to use simple KNN and SVD to build a dynamic recommender system.
A simple Steam Game recommender website. The recommendation result is dynamic, based on the game user puchased and what user click on the web page.
.
├── app.py
├── config
│ └── config.py
├── data
├── model //recommender models
├── natapp.sh // nat tool. You might select other nat tools
├── readme.md
├── static
│ ├── index.css
│ └── list.png
├── templates
│ └── index.html
├── user_init_rate // initial recommend list gerenate from user purchase history
├── user_item_rate // dynamic recommend list generated from user click
└── utils
└── web_utils.py
python app.py --temperature 0.5 --num_lines_load 4 --num_lines_init 8
# or flask run
--temperature
how much your recommender system react to the user click behavior.
--num_lines_load
how much lines of games to load when you click loadmore
--num_lines_init
how much lines of games to load at initial.
If you click like for one item, the rating for other similar items will be increased, such that they are more likely to be recommended. Click detail for one item will result in slightly increase on similar item score.
The design style is similar to steam official, click details to view details in steam official website.
user id update bar