Integrating Apache Mahout with AEM

Created byAnkit Gubrani / @ankitgubrani90

& Rima Mittal / @rimamittal

Speakers Bio

Agenda

• Introduction to Apache Mahout
• Machine Learning
• Recommendations
• AEM with Apache Mahout
• Demo
• Extension Points

Introduction to Apache Mahout

Introduction to Apache Mahout

• Project of the Apache Software Foundation.
• Producing free implementations of scalable machine learning algorithms, written in Java.

History

• Started as a Lucene sub-project.
• Became Apache TLP in April 2010.
• Latest version – 0.12.2 – Released on 13th June 2016.

Why Apache Mahout?

• Increasing volume of data!
• Traditional Data mining algorithms struggle to process very large datasets.
• Apache Mahout to the rescue!

Traditional Machine Learning

Machine Learning with Mahout

Applications

• Adobe, Facebook, LinkedIn, Twitter and Yahoo use Mahout internally.
• Twitter uses Mahout for interest modelling.
• Yahoo! Uses Mahout for pattern mining.

Machine Learning

Machine Learning

• Programming computers to optimize a Performance Criterion using Example Data or Past Experience
• Branch of Artificial Intelligence.
• Computers evolve behavior based on Empirical data.

Techniques

• Supervised Learning
• Use Labelled training data to create a classifier that can predict output for unseen inputs.
• Unsupervised Learning
• Use Unlabeled training data to create a function that can predict output.

Machine Learning with Apache Mahout

• Data Science use cases Mahout supports:
• Collaborative Filtering
• Clustering
• Classification

Collaborative Filtering

• User behavior mining to make product recommendations.

Clustering

• Organizing items into naturally occurring groups, such that items belonging to same group are similar to each other

Classification

• Learning from existing categorizations and assigning unclassified items to the best category

Recommendations

Apache Mahout Recommendation Engine

• Helps users find items they might like based on historical behavior and preferences.
• Mahout provides a rich set of components from which a customized recommender system can be constructed using a selection of Algorithms.

Architecture

• Top Level Packages
• DataModel
• UserSimilarity
• ItemSimilarity
• UserNeighboorhood
• Recommender

AEM with Apache Mahout

Checklist

• AEM 6.2
• Mahout as a Maven Dependency
  <dependency>
  <groupId>org.apache.mahout</groupId>
  <artifactId>mahout-mr</artifactId>
  <version>0.10.0</version>
  </dependency>
  

JCRDataModel

• DataModel
• Implementations representing a repository of information about users and their associated preferences.
• AbstractDataModel, JDBCDataModel, FileDataModel, GenericBooleanPrefDataModel, GenericDataModel.
• AEM - JCRDataModel.

Code (1/2)

• Using The AEM JCRDataModel

public JSONArray getUserBasedRecommendations(ResourceResolver resourceResolver, String userId, int numberOfRecommendations) {
//Creating JCRDataModel to fetch information from JCR
DataModel model = JCRDataModel.createDataModel(resourceResolver);
}

Code (2/2)

• AEM-Mahout Recommendation steps

UserSimilarity userSimilarity = getSimilarity(model);
UserNeighborhood neighborhood = getNeighbourHood(N_NEIGHOBUR_HOOD, userSimilarity, model);
GenericUserBasedRecommender recommender = new GenericUserBasedRecommender(model, neighborhood, userSimilarity);
recommendations = recommender.recommend(userIdHash, numberOfRecommedations, null, false);


AEM Product Recommendation

• User Based Recommendation
• Takes user ratings into consideration
• Based on PearsonCorrelationSimilarity
• Uses NearestNUserNeighborhood

Configuring JCRDataModel

• Configurations
• User Generated Content Path
• /content/usergenerated/asi/jcr
• Based on PearsonCorrelationSimilarity
• /etc/commerce/products/geometrixx-outdoors
• Uses NearestNUserNeighborhood
• Defaults to social/tally/components/response

Demo

Appendix

Appendix

• https://mahout.apache.org/
• http://www.slideshare.net/VaradMeru/introduction-to-mahout-and-machine-learning
• https://www.youtube.com/watch?v=iMAMYzfRiS4

Thank you.