recommender-algorithm-ALS

algorithm recommender_algorithm_ALS is

  input: my_ratings.txt 
  output: recommendations.txt
  
  // 1. Load & Parse Data
  
  my_ratings = rdd(parse(path/to/my_ratings))
  ratings = rdd(parse(path/to/small-ratings.csv))
  movies = rdd(parse(path/to/small-movies.csv))
  complete_ratings = rdd(parse(path/to/ratings.csv))
  complete_movies = rdd(parse(path/to/movies.csv))
  
  training = ratings.filter()
  validation = ratings.filter()
  test = ratings.filter()
  
  // 2. Train model & evaluate on validation set
  seed = 5L
  iterations = 10
  regularization_parameter = 0.1
  ranks = [4, 8, 12]
  errors = [0, 0, 0]
  err = 0
  tolerance = 0.02
  
  for rank in ranks
    model = ALS.train(training, rank, seed, iterations, regularization_parameter)
    predictions = model.PredictAll(validation).map()
  endfor
  
  bestRank = bestRank(model)
  
  training, test = complete_ratings.randomSplit([7, 3], seed=0L)
  complete_model = ALS.train(training, bestRank, seeds, iterations, regularization_parameter)
  predictions = complete_model.PredictAll(test).map()
  complete_ratings_data = complete_rating.filter().map()
  complete_data_with_new_ratings = rdd(complete_ratings_data.union(my_ratings))
  
  // 3. Make recommendations
  new_ratings_model = ALS.train(complete_data_with_new_ratings, best_rank, seed, iterations, regularization_parameter)
  recommendations = new_ratings_model.predictAll(complete_data_with_new_ratings)
  log(output, transformResults(recommendations))

end