Soluzioni Lab DIA: Recommendation con Python senza librerie

dia2022-00-recomm_nolibs.sols.py

# ESERCIZIO 1

# 1a
with open("items.csv", "r") as f:
    items = {int(iid): name for iid, name in csv.reader(f, delimiter=";")}

# 1b
len(items)

# 1c
items[2669]


# ESERCIZIO 2

# 2a
len(purchases)

# 2b
len(purchases) / len(users)


# ESERCIZIO 3

# 3a
purchases_by_item = {}
for uid, iid in purchases:
    purchases_by_item.setdefault(iid, set()).add(uid)

# 3b
max(map(len, purchases_by_item.values()))

# 3b
min(map(len, purchases_by_item.values()))


# ESERCIZIO 4

# 4
def user_similarity(uid1, uid2):
    """Count products purchased by both given users."""
    return len(purchases_by_user[uid1] & purchases_by_user[uid2])


# ESERCIZIO 5

# 5
def interest(uid, iid):
    """Estimate the interest of given user for given product."""
    return sum(
        user_similarities[(uid, ouid)]
        for ouid in purchases_by_item[iid]
        if uid != ouid
    )


# ESERCIZIO 6

# 6
max(
    interest_value
    for user_interests in interests_by_user.values()
    for interest_value in user_interests.values()
)


# ESERCIZIO 7

# 7a
suggestions_for_user_84 = set(iid for iid, score in sorted_interests_of_user_84[:N])

# 7b
def suggest(uid):
    """Recommend N products to given user."""
    interests = interests_by_user[uid].items()
    sorted_interests = sorted(interests, key=lambda x: x[1], reverse=True)
    return set(iid for iid, score in sorted_interests[:N])


# ESERCIZIO 8

# 8a
new_purchases_by_user = {}
for uid, iid in new_purchases:
    new_purchases_by_user.setdefault(uid, set()).add(iid)

# 8b
max(map(len, new_purchases_by_user.values()))

# 8b
sum(map(len, new_purchases_by_user.values())) / len(users)


# SELEZIONE CASUALE

def suggest_random(uid):
    unpurchased_iids = list(interests_by_user[uid].keys())
    suggested_iids = random.sample(unpurchased_iids, N)
    return set(suggested_iids)

random_suggestions_by_user = {uid: suggest_random(uid) for uid in users.keys()}

randomly_satisfied_users = {uid for uid in users.keys()
                            if random_suggestions_by_user[uid] & new_purchases_by_user.get(uid, set())}

len(randomly_satisfied_users) / len(users)

results = []
for n in range(1000):
    random.seed(n)
    random_suggestions_by_user = {uid: suggest_random(uid) for uid in users.keys()}
    randomly_satisfied_users = {uid for uid in users.keys()
                                if random_suggestions_by_user[uid] & new_purchases_by_user.get(uid, set())}
    results.append(len(randomly_satisfied_users) / len(users))

sum(results) / len(results)