[an error occurred while processing this directive]
<div class="title">Project D: Music Recommendation</div>
<p><b>performed by Group 3</b></p>

<p>
Recommending interesting music to people is at the core of modern music players. Based on existing ratings and similarity between users, recommended songs have a high chance of being to the taste of a user. Your task is to apply a state-of-the-art graph-based recommendation algorithm to a <a href="http://webscope.sandbox.yahoo.com/catalog.php?datatype=r&amp;did=2">dataset from Yahoo! Music</a>.
</p>

<div class="post_title">Idea</div>
<ul>
<li>Run the Alternating Least Squares (ALS) algorithm</li>
<li>Convert latent vector into actual recommendations</li>
<li>Determine the quality of prediction by comparing the computed predictions with the test set</li>

<li><i>Stretch goals: Map ids back to actual songs, build interface where one can select their favorite songs and gets recommendations for others</i></li>
</ul>

<div class="post_title">Technology</div>
<p>
This analysis will be using the <a href="http://grafos.ml/okapi.html">Okapi Collaborative Filtering</a> library on top of 
<a href="http://giraph.apache.org">Giraph</a>. 
</p>

<div class="post_title">Data</div>
<p>
The music interest data is available in text format. The data will be available on HDFS at <code>/user/hannesm/lsde/music</code> . Example:
</p>
<p>Music interest data (user, song, rating)</p>

<pre style="font-size: 9px; white-space: pre-wrap;">
49      2169    5
49      2180    2
50      311     5
50      325     1
</pre>

<div class="post_title">Literature</div>
<ul>
<li><a href="http://www.hpl.hp.com/personal/Robert_Schreiber/papers/2008%20AAIM%20Netflix/netflix_aaim08%28submitted%29.pdf">Large-Scale Parallel Collaborative Filtering for the Netflix Prize</a></li>
<li><a href="http://kowshik.github.io/JPregel/pregel_paper.pdf">Pregel: a system for large-scale graph processing</a></li>
</ul>
[an error occurred while processing this directive]