Install dependencies from requirements.txt.

• There is a directory called preprocessing. It is used to clean & preprocess the data.
• The second directory is processing. It contains all the AI algorithms.
• The third important directory is datasets. And it contains the csv files.
• All the other directories are of Django.

• Open the root directory of project in cmd.
• Then run python preprocessing/Preprocessing.py. It will generate a new file called preprocessed.csv in datasets directory.
• Now run the server using command python manage.py runserver.
• Now open localhost link in browser with a port number given in cmd.
• Now click on any video. A new page with that video will open.
• Now scroll down a little bit. And you will see recommended videos.
• You can also search videos. Or click on a particular tag to see recommendations.

The dataset used is available here kaggle.com/datasnaek/youtube-new.

There are 16 different columns. Including video title, description, tags, likes, dislikes, number of comments etc.

There are over 40000 rows. But actually many rows are duplicated. Unique rows in the dataset is about 6000.

• In preprocessing we use different techniques to clean the data. Like converting all textual data to lower case. Removing all special characters etc.
• And in recommendation we use cosine similarity algorithm to match the similar word in tags of different videos.