ContentBasedFiltering.py

import numpy as np
from sklearn.preprocessing import MultiLabelBinarizer, normalize
import faiss

class ContentBasedFiltering:
    def __init__(self, games):
        self.games = games
        self.index = None
        self._prepare_data()

    def _prepare_data(self):
        mlb_genres = MultiLabelBinarizer()
        mlb_categories = MultiLabelBinarizer()
        self.games["genres_list"] = self.games["genres"].apply(lambda x: x.split(";") if isinstance(x, str) else [])
        genres_encoded = mlb_genres.fit_transform(self.games["genres"])
        categories_encoded = mlb_categories.fit_transform(self.games["categories"])
        self.games["developer_encoded"] = self.games["developer"].astype("category").cat.codes
        self.games["publisher_encoded"] = self.games["publisher"].astype("category").cat.codes
        game_features = np.hstack([genres_encoded, categories_encoded, self.games[["developer_encoded", "publisher_encoded"]].values])
        game_features = game_features.astype(np.float32)
        normalize(game_features, norm="l2")
        self.index = faiss.IndexFlatIP(game_features.shape[1])
        faiss.normalize_L2(game_features)
        self.index.add(game_features)

    def find_similar_games(self, game_name, top_n=5):
        game_idx = self.games.index[self.games["name"] == game_name].tolist()[0]
        distances, indices = self.index.search(self.index.reconstruct(game_idx).reshape(1, -1), top_n + 1)
        return self.games.iloc[indices[0][1:]][["name", "appid"]].values.tolist()