diff --git a/dags/business/business/community/consumption.py b/dags/business/business/community/consumption.py index b9a0b87..f392ac0 100644 --- a/dags/business/business/community/consumption.py +++ b/dags/business/business/community/consumption.py @@ -30,6 +30,19 @@ reasonable_df = pd.read_excel(path + '/财力评级分布情况20221215.xlsx') reasonable_dict = reasonable_df.set_index('city')['是否合理'].to_dict() +# Min-Max 归一化到 [1, 100] +def min_max_scale(df, column, new_min=1, new_max=100): + X_min = df[column].min() + X_max = df[column].max() + + # 防止分母为零(如果所有值相等) + if X_max == X_min: + df['consumption_score'] = new_min # 如果所有值相同,设为下界 + else: + df['consumption_score'] = (new_min + (df[column] - X_min) * (new_max - new_min) / (X_max - X_min)) + df['consumption_score'] = df['consumption_score'].astype(int) + return df + def calc_consumption_potential(city): """ 计算消费潜力 @@ -107,7 +120,14 @@ def calc_consumption_potential(city): def gmm_train(total_rate, consumption_df, source, target): # 高斯混合模型 - gmm = GaussianMixture(n_components=total_rate, covariance_type='tied', max_iter=200, n_init=10).fit(consumption_df[[source]]) + gmm = GaussianMixture( + n_components=total_rate, + covariance_type='full', # 更灵活的协方差矩阵 + max_iter=200, + n_init=10, + reg_covar=1e-6, # 正则化防止奇异矩阵 + random_state=42 + ).fit(consumption_df[[source]]) consumption_df.loc[:, 'cluster'] = gmm.predict(consumption_df[[source]]) consumption_df.loc[:, 'price_mean_cluster'] = consumption_df.groupby('cluster')[source].transform('mean') consumption_df.loc[:, target] = consumption_df['price_mean_cluster'].rank(method='dense', ascending=False).astype(int) @@ -145,22 +165,18 @@ def calc_consumption_rating(community_summary_df): df = gmm_train(total_rate, df, "relative_price_ranking_score", "relative_price_rating") df = gmm_train(total_rate, df, "avg_cost_score", "avg_cost_rating") - # 总分数 - total_score = 100 - if len(df) < total_score: - # 对于样本低于5的直接排名 - df["consumption_score"] = df['consumption_potential'].rank(method='dense', ascending=False).astype(int) - else: - # 极值设置 - df = filter_min_max(df) - df = gmm_train(total_score, df, source, "consumption_score") + # 填充缺失值 + df["consumption_rating"] = df["consumption_rating"].fillna(df["total_price_rating"]) + + # 应用归一化 + df = min_max_scale(df, 'consumption_potential') # 批量更新 for index, item in df.iterrows(): row = dict() _id = str(item["_id"]) row['consumption_rating'] = item['consumption_rating'] - row['consumption_score'] = 101 - item['consumption_score'] if pd.notna(item['consumption_score']) else None + row['consumption_score'] = item['consumption_score'] if pd.notna(item['consumption_score']) else None row['consumption_potential'] = item['consumption_potential'] row['m_date'] = dateutils.iso_format() row['sub_factor'] = { diff --git a/dags/business/business/community/consumption_stat.py b/dags/business/business/community/consumption_stat.py index c298e69..b63ade1 100644 --- a/dags/business/business/community/consumption_stat.py +++ b/dags/business/business/community/consumption_stat.py @@ -165,7 +165,7 @@ def detail(city): "获取方式", "相对价格排名", "人均消费", "物业费", "物业费获取方式", '置信度']) nextcloudutils.up_to_nextcloud(save_png_path + "/" + aim_file, nextcloud_aim_file_path + "/" + aim_file) - dingutils.send_text("【统计】财力评级统计完毕,请点击 https://nextcloud.fangjia.com/s/nMbzk6ZKJDimapf 查看", "32400b949005817550937f82a82c064fe5be65c23204306442c15e8b485d7ef2") + #dingutils.send_text("【统计】财力评级统计完毕,请点击 https://nextcloud.fangjia.com/s/nMbzk6ZKJDimapf 查看", "32400b949005817550937f82a82c064fe5be65c23204306442c15e8b485d7ef2") sys_city = sys.argv[1] if len(sys.argv) > 1 else None