In this project, we tried to recommend movies to users based on their liked activity as well as the liked activity of their friends. We used Apache Mahout for the Machine Learning Algorithms and Graph API explorer to access Facebook activity by creating a Facebook App.
(NEHA) Call Girls Katra Call Now 8617697112 Katra Escorts 24x7
Movie recommendation system using Apache Mahout and Facebook APIs
1. Movie Recommendation System using
Apache Mahout with Facebook
• Recommendation systems are widely used as they provide assistance in decision making.
• Problem Statement: Using popular Machine Learning algorithms, provide movie recommendations
based on user and friends’ movie likes on Facebook.
• Apache Mahout: scalable machine learning library implemented on top of Apache Hadoop.
• Development Environment: Apache Mahout on Eclipse with Maven integration
• Data Set: Graph API Explorer, Movie Lens dataset
• Algorithm used: Collaborative Filtering User-based/Item Based recommendation identify users
by similar preferences (movie likes)
• Pearson Correlation
• Log Likelihood
• Nearest Neighbour
3. Conclusion and Future Work
Challenges
Change in Facebook permissions – information of only select few friends could be retrieved.
Old movie data set from Movie Lens Database.
Lessons learnt:
Limitations of Pearson Correlation – Only those users who declare a preference are considered i.e.
only those users who have liked a movie are considered from the sample size.
Log Likelihood performs better at finding similar users than Pearson Correlation.
Facebook application development and using Graph API Explorer.
Real time applications of statistical methods – chi square test, hypothesis testing in understanding the
implementation of machine learning algorithms.
Future Work:
Improve User Experience
Improve recommendation accuracy
Twitter – use Big Data and Hadoop clusters