MOVIE_RATING_PATH = "../../data/movie_rating_20190219_1.csv"

# -----------------------------基于用户的协同过滤-------------------------------


def eu_distance(user1, user2):
    # 用户间距离：欧氏距离
    distance = 0
    cal = 0
    for user1_key in user1.keys():
        if user1_key in user2.keys():
            distance = distance + pow(user1[user1_key] - user2[user1_key], 2)
            cal = cal + 1
    return (distance ** 0.5) / (cal + 0.001)


def cos_distance(user1, user2):
    # 用户间距离：余弦距离
    distance = 0
    user1_norm = 0
    user2_norm = 0
    cal = 0
    for user1_key in user1.keys():
        if user1_key in user2.keys():
            distance = distance + user1[user1_key] * user2[user1_key]
            user1_norm = user1_norm + user1[user1_key] * user1[user1_key]
            user2_norm = user2_norm + user2[user1_key] * user2[user1_key]
            cal = cal + 1
    res = distance / ((user1_norm ** 0.5) * (user2_norm ** 0.5) + 0.001)
    return res


def top_similar(data, user, num=10):
    # 最相似的N个用户
    res = []
    for userid in data.keys():
        if userid == user:
            continue
        # sim = eu_distance(data[user], data[userid])
        sim = cos_distance(data[user], data[userid])
        res.append((userid, sim))
    res.sort(key=lambda val: val[1])
    print(res[:num])
    return res[:num]


def recommend(data, user, user_num=10, rec_num=10):
    # 进行推荐的主函数
    # 获得最接近的几个用户
    user_close = top_similar(data, user, num=user_num)
    # 获取这些用户看过的所有item
    movie = {}
    for item in user_close:
        for user_close_movie in data[item[0]]:
            if user_close_movie not in movie:
                movie[user_close_movie] = []
            movie[user_close_movie].append(
                (item[1], data[item[0]][user_close_movie]))
    # 给所有相关item进行打分
    movie_cal = []
    for item in movie:
        score = 0
        weight = 0
        for cal in movie[item]:
            score = score + cal[1]
            weight = weight + 1
        movie_cal.append((item, score / weight))
    movie_cal.sort(key=lambda val: val[1], reverse=True)
    # 避免推荐用户已经看过的item
    res = []
    USER = data[user]
    for item in movie_cal:
        if item[0] in USER:
            continue
        res.append(item)
        if len(res) >= rec_num:
            break
    if len(res) < rec_num:
        print("合适item不足目标数量")
        return res
    return res[:rec_num]

# -----------------------------基于用户的协同过滤-------------------------------


# json格式化-user-movie-rating-加载版本
data = {} # DIC用用户-item-打分的形式
with open(MOVIE_RATING_PATH, 'r', encoding='UTF-8') as f:
    idx = 0
    for line in f:
        if idx == 0:
            idx = 1
            continue
        ll = line.strip().split(",")
        if ll[0] not in data:
            data[ll[0]] = {}
        data[ll[0]][ll[3]] = float(ll[1])

print("data imported: %s" % len(data))

print(recommend(data, "1"))