BrandCultivation/models/recall/itemCF/similarity_matrix.py

from database import MySqlDao
from itertools import combinations
from models.rank.data.config import OrderConfig
import numpy as np
import pandas as pd
from tqdm import tqdm


class SimilarityMatrix:
    def __init__(self, data):
        self._order_data = data.copy()
        self._load_data()
        self._build_co_occurace_matrix()
        
    def _load_data(self):
        """加载订单记录表"""
        self._order_data =self._order_data[OrderConfig.FEATURE_COLUMNS]
        
        # 数据清洗
        self._order_data["sale_qty"] = self._order_data["sale_qty"].fillna(0)
        self._order_data = self._order_data.groupby(["cust_code", "product_code"], as_index=False)["sale_qty"].sum()
        self._order_data = self._order_data[self._order_data["sale_qty"] != 0]
        
    def _build_co_occurace_matrix(self):
        """构建商户共现矩阵"""
       # 获取所有商户的唯一列表
        self._shops = self._order_data["cust_code"].unique()
        num_shops = len(self._shops)
        
        # 创建商户到索引的映射
        self._shops_to_index = {shop: idx for idx, shop in enumerate(self._shops)}
        # 初始化共现矩阵（上三角部分）
        self._co_occurrence_matrix = np.zeros((num_shops, num_shops), dtype=int)
        
        # 按照品规分组
        grouped = self._order_data.groupby("product_code")["cust_code"].apply(list)
        
        # 遍历每个品规的商户列表
        for shop_in_product in tqdm(grouped, desc="正在构建共现矩阵..."):
            # 生成商户对
            shop_pairs = combinations(shop_in_product, 2)
            for shop1, shop2 in shop_pairs:
                # 获取商户索引
                idx1 = self._shops_to_index[shop1]
                idx2 = self._shops_to_index[shop2]
                # 更新共现矩阵
                self._co_occurrence_matrix[idx1, idx2] += 1
                self._co_occurrence_matrix[idx2, idx1] += 1
                
    def generate_similarity_matrix(self):
        """使用向量计算商铺之间的相似度矩阵"""
        # 计算每个商铺售卖品规的总次数
        shop_counts = self._order_data.groupby("cust_code").size()
        
        # 将商户售卖次数转换为数组
        counts = np.array([shop_counts[shop] for shop in self._shops_to_index.keys()])
        
        # 计算分母部分 (sqrt(count_i * count_j))
        denominator = np.sqrt(np.outer(counts, counts))
        
        # 计算相似度矩阵
        self._similarity_matrix = self._co_occurrence_matrix / denominator
        
        # 将对角线设置为1
        np.fill_diagonal(self._similarity_matrix, 1.0)
        
        # 保存结果
        self._similarity_matrix = pd.DataFrame(self._similarity_matrix, index=self._shops, columns=self._shops)
        # self._similarity_matrix.to_csv(save_path, index=True, encoding="utf-8")
        return self._similarity_matrix
        
if __name__ == "__main__":
    similarity_matrix_save_path = "./data/itemcf/similarity.csv"
    similarity_matrix = SimilarityMatrix("00000000000000000000000011445301")
    similarity_matrix.generate_similarity_matrix(similarity_matrix_save_path)