Ch2 Ex3-Programming Collective Intelligence

import recommendations as recommendations

def getn_similarcritics(prefs,person,n=5):
    dict1 = {}
    high = 0
    # Compare critics to person and add to dict1
    for b in prefs:
        if b != person:
            crit2 = str(b)
            pear = recommendations.sim_pearson(prefs, person, crit2)
            if pear not in dict1:
                dict1[pear] = [person, crit2]

    # return top n critics
    res = dict1.keys()
    res.sort()
    res.reverse()
    listout = []
    for a in range(n):
        listout.append(dict1[res[a]][1])

    return listout


def getRecommendations1(prefs, person, similarity=recommendations.sim_pearson):
    totals = {}
    simSums = {}

    others = getn_similarcritics(prefs,person)
    for other in others:
        sim = similarity(prefs, person, other)

        # ignore scores of zero or lower
        if sim <= 0: continue
        for item in prefs[other]:

            # only score movies I haven't seen yet
            if item not in prefs[person] or prefs[person][item] == 0:

                # Similarity * Score
                totals.setdefault(item,0)
                totals[item] += prefs[other][item]*sim

                # Sum of similarities
                simSums.setdefault(item,0)
                simSums[item] += sim

    # Create the normalized list
    rankings = [(total/simSums[item],item) for item, total in totals.items()]

    # Return the sorted list
    rankings.sort()
    rankings.reverse()
    return rankings


if __name__ == "__main__":
    prefs = recommendations.loadMovieLens()
    print getRecommendations1(prefs, '87')[0:30]


""" User-based efficiency. The user-based filtering algorithm is inefficient because it compares a user to all other users every time a recommendation is needed. Write a function to precompute user similarities, and alter the recommendation code to use only the top five other users to get recomendations

person is the person to get recommendations for
prefs  is the dataset
n      is the number of critics to get recommendations from

"""