Youtube Video Categorization

The repository describes Machine learning methods used to implement a Youtube Video Categorization model which gives a caregory to a video given it's title or description.

OBJECTIVE

Classifying videos (based on title and description) into the following categories:

1. Travel
2. Tutorial
3. Science
4. Arts
5. History
6. Movies
7. Vlog
8. Food
9. Music
10. Fashion

Machine Learning algorithms used:

1. K-Nearest Neighbours
2. Logistic Regression
3. Naive Bayes Classifier
4. Decision Tree Classifier
5. Random Forests Classifier
6. Support Vector Classifier
7. Ada Boost
8. XGBoost Classifer

Results

Algorithm	Hyperparameter	Accuracy
Naive Bayes	Alpha - 10	86.5
Logistic Regression	0.01, L2	88.67
KNN	K = 43	83.4
Random Forests	Tf-IDF	87.69
XGBoost	Lemmatized	89.97
Gradient Boost	Tf-Idf	88.01
SVC	Linear , 'C' = 1	88.6

EDA

I tried the T-SNE algorithm for finding the lower dimensional representation of the data.

PCA decomposition