fariha23 / youtube_data_analysis_video_categories

YouTube_Data_Analysis_Video_Categories

Title: Predicting_Popularity_of_YouTube_Videos

Author: Fariha Baloch, PhD

This repository contains code and reports on project titled "Predicting Popularity of YouTube Videos". Using YouTube data from Kaggle for the USA and Mexico several data cleaning and data analysis steps were performed and as final step using Machine Learning methods, popularity of the videos was predicted

Data Wrangling was performed as the first step in the process - where data was converted to dataframe and cleaned up to be used in analysis

Data Story was performed on the data obtained from Data Wrangling step. Here several graphs were formed to explain how the data is layed out. Where the unique videos are popular and where majority of the views exist, etc.

Statistical Analysis was performed on the data obtained in last steps. Questions like the distributions of the data and feature dependence on each other was performed. Using Bayseian Inference the time period of a high views and low views on a single video was performed.

Machine Learing was next. Used clustering and classification methods to perform ML on the data. Using clustering algorithm KMeans, three popularity classes were predicted. Using KNN, RF, GB and AdaBoost algorithms on the data with more features added, perdiction on popularity was performed.

Individual reports of each step mentioned above are provided in this repository with titled folders

See the Capstone 1 Report and Presentation for more comprehensiove details on the project steps