Personalized, Weighted Recommendations - Rank All Content

问题

I'm building a social site for musicians. I would like to take the whole list of songs from the database and rank each one based on its relevancy to the logged in user. One reason for this is that I'd like there to always be 10 songs recommended, even if the user signed up 45 seconds ago.

The factors I'm using are:

The band members of songs (all would be members of the site, might have all quit the song)
The logged in user's member connections (may be none)
The most recent update in the song (will at least be the day the song was "created")
The (sub)genre of the song (will always be set)
The (sub)genre of the user (will always be set)

The (sub)genres belong to more general genres, so I figure I can either weight a song higher if the logged in user and the song's (sub)genre are the same, or a little less higher if they at least belong to the same general genre.

I'm not sure if I should be aiming for an Apriori or Bayesian algorithm... It seems to be somewhere in between. I have a couple of lines borrowed from explanations of the Hacker News algorithm that I've tweaked. I'm not quite sure how to put the final results together to get my ranking.

Here's the code I've written so far (for all intents and purposes, a studio is a song). I know queries in loops like this are never a good idea, but I've done it this way because there's some memcaching schemes I use and this will probably also be batched:

function studio_relevance($user_id, $user_genre) {
    global $cxn;

$query = "SELECT * FROM studio WHERE project_status='active'";
$result = mysqli_query($cxn, $query) or die($query.': '.mysqli_error($cxn));

$data = array();

while ($studio = mysqli_fetch_object($result)) {
    //Find similarities in social connections and band
    $query_b = "SELECT * FROM band WHERE studio_id=".$studio->studio_id;
    $result_b = mysqli_query($cxn, $query_b) or die($query_b.': '.mysqli_error($cxn));

    $the_band = array();
    while ($people = mysqli_fetch_array($result_b, MYSQLI_ASSOC)) {
        $the_band[] = $people['user_id'];
    }

    $studio_band_count = count($the_band);

    $query = "SELECT * FROM idols WHERE friend_id=".$user_id;
    $result_b = mysqli_query($cxn, $query_b) or die($query_b.': '.mysqli_error($cxn));

    $idol = array();
    while ($people = mysqli_fetch_array($result_b, MYSQLI_ASSOC)) {
        $idol[] = $people['artist_id'];
    }

    $same_band = array_intersect($the_band, $idol);
    $same_band_count = count($same_band);

    ########
    $similar_band = $same_band_count / $studio_band_count;
    ########


    //Find the most recent activity
    $query_b = "SELECT * FROM studio_feed WHERE studio_id=".$studio->studio_id." ORDER BY feed_id DESC LIMIT 1";
    $result_b = mysqli_query($cxn, $query_b) or die($query_b.': '.mysqli_error($cxn));
    $tab = mysqli_fetch_object($result_b);

    $time_diff = strtotime('now') - strtotime($tab->timestamp);
    $hours = $time_diff / 3600;

    ########
    $last_activity = pow(($hours+2), 1.8);
    ########


    //Compare genres
    $genre_weight = 1;

    if ($studio->songGenre == $user_genre) {
        $genre_weight = 3;
    }
    else {
        $query_b = "SELECT * FROM genres";
        $result_b = mysqli_query($cxn, $query_b) or die($query_b.': '.mysqli_error($cxn));

        $genres = array();
        $user_genre_cat = 0;

        while ($genre = mysqli_fetch_object($result_b)) {
            $genres[$genre->genre_cat][] = $genre->genre_id;

            if ($genre->genre_id == $user_genre) {
                $user_genre_cat = $genre->genre_cat;
            }
        } 

        if (in_array($studio->songGenre, $genres[$celeb_cat])) {
            $genre_weight = 2;
        }
    }


    //Find final result
    //$final = $similar_band + $last_activity + $genre_weight;
    //$hours / pow(($similar_band+2), $genre_weight);
}

return $data;
}

Any suggestions would be greatly appreciated! Am I on the right track or going about this all wrong?

回答1:

tl;dr. I'd advise you to rewrite your question from scratch, or check out the CodeReview site. https://codereview.stackexchange.com/

来源：https://stackoverflow.com/questions/9097919/personalized-weighted-recommendations-rank-all-content

标签

php

algorithm

recommendation-engine