{"id":7730,"date":"2026-03-17T10:50:09","date_gmt":"2026-03-17T01:50:09","guid":{"rendered":"https:\/\/blog.since2020.jp\/?p=7730"},"modified":"2026-03-17T10:50:53","modified_gmt":"2026-03-17T01:50:53","slug":"bigquery-ml-uplift-modeling","status":"publish","type":"post","link":"https:\/\/since2020.jp\/media\/bigquery-ml-uplift-modeling\/","title":{"rendered":"BigQuery ML\u3067\u4f5c\u308bUplift Modeling"},"content":{"rendered":"\n

\u3080\u3084\u307f\u306a\u30af\u30fc\u30dd\u30f3\u914d\u5e03\u3067\u5229\u76ca\u3092\u524a\u3063\u3066\u3044\u307e\u305b\u3093\u304b\uff1f\u56e0\u679c\u63a8\u8ad6\u300cUplift Modeling\u300d\u3092\u7528\u3044\u3066\u300c\u30af\u30fc\u30dd\u30f3\u3092\u914d\u308b\u3079\u304d\u9867\u5ba2\u300d\u3060\u3051\u3092\u7279\u5b9a\u3057\u3001\u671f\u5f85\u5229\u76ca(ROI)\u3092\u6700\u5927\u5316\u3059\u308b\u624b\u6cd5\u3092\u5b9f\u8df5\u89e3\u8aac\u3057\u307e\u3059\u3002BigQuery ML\u306e\u5236\u7d04\u3092\u30cf\u30c3\u30af\u3059\u308b\u300c\u7269\u7406\u7684\u30ea\u30b5\u30f3\u30d7\u30ea\u30f3\u30b0\u306b\u3088\u308b\u7591\u4f3cIPW\u300d\u306e\u5b9f\u88c5\u30c6\u30af\u30cb\u30c3\u30af\u3082\u5fc5\u898b\u3067\u3059\u3002<\/p>\n\n\n

\u30c7\u30fc\u30bf\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u3068\u7279\u5fb4\u91cf\u30a8\u30f3\u30b8\u30cb\u30a2\u30ea\u30f3\u30b0<\/h2>\n

\u30e2\u30c7\u30eb\u306e\u4e88\u6e2c\u7cbe\u5ea6\u306f\u30c7\u30fc\u30bf\u306e\u524d\u51e6\u7406\u3067\u6c7a\u307e\u308b\u3068\u8a00\u3063\u3066\u3082\u904e\u8a00\u3067\u306f\u3042\u308a\u307e\u305b\u3093\u3002\u4eca\u56de\u306e\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u3067\u306f\u3001\u751f\u30c7\u30fc\u30bf\u304b\u3089\u6bb5\u968e\u7684\u306b\u30c7\u30fc\u30bf\u30de\u30fc\u30c8\u3092\u69cb\u7bc9\u3059\u308b\u30b9\u30bf\u30fc\u30b9\u30ad\u30fc\u30de\u69cb\u9020\u3092\u63a1\u7528\u3057\u307e\u3059\u3002<\/p>\r\n

\uff11\uff0eSTG\u5c64\uff08Staging\uff09<\/strong>: \u30de\u30b9\u30bf\u30c7\u30fc\u30bf\u3068\u30c8\u30e9\u30f3\u30b6\u30af\u30b7\u30e7\u30f3\u306e\u30af\u30ec\u30f3\u30b8\u30f3\u30b0<\/p>\r\n

\uff12\uff0eINT\u5c64\uff08Intermediate\uff09<\/strong>: \u30d3\u30b8\u30cd\u30b9\u30ed\u30b8\u30c3\u30af\u306e\u9069\u7528\uff08LTV\u30d1\u30fc\u30bb\u30f3\u30bf\u30a4\u30eb\u3001RFM\u3001\u30ea\u30b9\u30af\u30d5\u30e9\u30b0\u306e\u7b97\u51fa\uff09<\/p>\r\n

\uff13\uff0eMART\u5c64\uff08Mart\uff09<\/strong>: \u6a5f\u68b0\u5b66\u7fd2\u7528\u30fbBI\u53ef\u8996\u5316\u7528\u306e\u6700\u7d42\u30c6\u30fc\u30d6\u30eb<\/p>\r\n

\u7279\u306b\u3001\u5358\u306a\u308b\u5c5e\u6027\u30c7\u30fc\u30bf\u3060\u3051\u3067\u306a\u304f\u3001\u300c\u8cfc\u8cb7\u505c\u6ede\u30ea\u30b9\u30af\u300d\u300c\u5546\u54c1\u306e\u6b20\u54c1\u30ea\u30b9\u30af\u300d\u3068\u3044\u3063\u305f\u884c\u52d5\u30d9\u30fc\u30b9\u306e\u52d5\u7684\u7279\u5fb4\u91cf\u3092\u7d44\u307f\u8fbc\u3093\u3067\u3044\u308b\u70b9\u304c\u30dd\u30a4\u30f3\u30c8\u3067\u3059\u3002<\/p>\r\n

<\/p>\n\n

Uplift Modeling\u306e\u5b9f\u88c5\uff1aT-Learner\u3068\u7591\u4f3cIPW\u306e\u30cf\u30c3\u30af<\/h2>\n\u306a\u305c\u3042\u3048\u3066BigQuery ML\uff08BQML\uff09\u3067\u5b9f\u88c5\u3059\u308b\u306e\u304b\uff1f<\/b>\r\n

Python\u3092\u7528\u3044\u305fVertex AI\u74b0\u5883\u3084\u30ed\u30fc\u30ab\u30eb\u306eJupyter Notebook\u306a\u3069\u3001\u67d4\u8edf\u306a\u30e2\u30c7\u30ea\u30f3\u30b0\u74b0\u5883\u304c\u3042\u308b\u4e2d\u3067\u3001\u306a\u305c\u3042\u3048\u3066BQML\u3092\u9078\u629e\u3059\u308b\u306e\u3067\u3057\u3087\u3046\u304b\u3002\u305d\u308c\u306f\u300cData Gravity\uff08\u30c7\u30fc\u30bf\u306e\u91cd\u529b\uff09\u300d\u3068\u300cMLOps\u306e\u5bb9\u6613\u3055\u300d\u306b\u5727\u5012\u7684\u306a\u30e1\u30ea\u30c3\u30c8\u304c\u3042\u308b\u304b\u3089\u3067\u3059\u3002<\/p>\r\n

\u30fb\u30c7\u30fc\u30bf\u79fb\u52d5\u30b3\u30b9\u30c8\u3068\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u30ea\u30b9\u30af\u306e\u6392\u9664<\/strong>: \u6570\u5104\u884c\u306b\u53ca\u3076\u30c8\u30e9\u30f3\u30b6\u30af\u30b7\u30e7\u30f3\u3084\u30ed\u30b0\u30c7\u30fc\u30bf\u306f\u3001\u3059\u3067\u306bBigQuery\uff08\u30c7\u30fc\u30bf\u30a6\u30a7\u30a2\u30cf\u30a6\u30b9\uff09\u5185\u306b\u5b58\u5728\u3057\u307e\u3059\u3002\u3053\u308c\u3092\u6a5f\u68b0\u5b66\u7fd2\u306e\u305f\u3081\u306b\u5916\u90e8\u74b0\u5883\u3078\u30a8\u30af\u30b9\u30dd\u30fc\u30c8\u3059\u308b\u3053\u3068\u306f\u3001\u901a\u4fe1\u30b3\u30b9\u30c8\u306e\u5897\u5927\u3084\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u30ea\u30b9\u30af\u306b\u7e4b\u304c\u308a\u307e\u3059\u3002BQML\u306a\u3089\u3001\u30c7\u30fc\u30bf\u304c\u7f6e\u304b\u308c\u305f\u5834\u6240\u3067\u305d\u306e\u307e\u307e\u5b66\u7fd2\u30fb\u63a8\u8ad6\u304c\u53ef\u80fd\u3067\u3059\u3002<\/p>\r\n

\u30fb\u30c7\u30fc\u30bf\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u3068\u306e\u30b7\u30fc\u30e0\u30ec\u30b9\u306a\u7d71\u5408<\/strong>: dbt\u3084\u30b9\u30b1\u30b8\u30e5\u30fc\u30eb\u3055\u308c\u305f\u30af\u30a8\u30ea\uff08Scheduled Queries\uff09\u306a\u3069\u3001\u65e2\u5b58\u306eSQL\u30d9\u30fc\u30b9\u306e\u30c7\u30fc\u30bf\u51e6\u7406\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u306e\u5ef6\u9577\u7dda\u4e0a\u306b\u3001\u300cML\u306e\u5b66\u7fd2\u30fb\u63a8\u8ad6\u300d\u3092\u7d44\u307f\u8fbc\u3080\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\r\n

\u30fb\u904b\u7528\u304a\u3088\u3073\u4fdd\u5b88\u306e\u5c5e\u4eba\u5316\u9632\u6b62<\/strong>: \u5225\u9014\u63a8\u8ad6\u7528\u306eAPI\u30b5\u30fc\u30d0\u30fc\u306a\u3069\u3092\u7acb\u3066\u308b\u5fc5\u8981\u304c\u306a\u304f\u3001SQL\u3092\u7406\u89e3\u3067\u304d\u308b\u30c7\u30fc\u30bf\u30a8\u30f3\u30b8\u30cb\u30a2\u3084\u30a2\u30ca\u30ea\u30b9\u30c8\u3067\u3042\u308c\u3070\u3001\u30e2\u30c7\u30eb\u306e\u4fdd\u5b88\u904b\u7528\u3092\u5bb9\u6613\u306b\u5f15\u304d\u7d99\u3050\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\r\n

\u3053\u306e\u3088\u3046\u306a\u30d3\u30b8\u30cd\u30b9\u904b\u7528\u4e0a\u306e\u5f37\u529b\u306a\u30e1\u30ea\u30c3\u30c8\u304c\u3042\u308b\u304b\u3089\u3053\u305d\u3001\u30c7\u30fc\u30bf\u30a6\u30a7\u30a2\u30cf\u30a6\u30b9\u5185\u3067\u5b8c\u7d50\u3059\u308bBQML\u306e\u6d3b\u7528\u304c\u30e2\u30c0\u30f3\u306a\u30c7\u30fc\u30bf\u57fa\u76e4\u3067\u597d\u307e\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\r\nT-Learner\u3068IPW\uff08\u50be\u5411\u30b9\u30b3\u30a2\u91cd\u307f\u4ed8\u3051\uff09<\/b>\r\n

Uplift Modeling\u306e\u4ee3\u8868\u7684\u306a\u624b\u6cd5\u306b\u3001\u4ecb\u5165\uff08\u30af\u30fc\u30dd\u30f3\u914d\u5e03\uff09\u3042\u308a\u30fb\u306a\u3057\u306e2\u3064\u306e\u30e2\u30c7\u30eb\u3092\u69cb\u7bc9\u3059\u308b\u300cT-Learner\u300d\u304c\u3042\u308a\u307e\u3059\u3002\u3057\u304b\u3057\u3001\u904e\u53bb\u306e\u89b3\u6e2c\u30c7\u30fc\u30bf\u306b\u306f\u300c\u30af\u30fc\u30dd\u30f3\u3092\u3082\u3089\u3044\u3084\u3059\u3044\u4eba\u30fb\u3082\u3089\u3044\u306b\u304f\u3044\u4eba\u300d\u3068\u3044\u3046\u9078\u629e\u30d0\u30a4\u30a2\u30b9\uff08Selection Bias\uff09\u304c\u542b\u307e\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\r\n

\u3053\u308c\u3092\u88dc\u6b63\u3059\u308b\u305f\u3081\u306b\u3001\u50be\u5411\u30b9\u30b3\u30a2\uff08Propensity Score\uff09\u306e\u9006\u6570\u3067\u91cd\u307f\u4ed8\u3051\u3092\u884c\u3046 IPW\uff08Inverse Probability Weighting\uff09 \u3092\u9069\u7528\u3057\u307e\u3059\u3002\u6570\u5f0f\u3067\u8868\u3059\u3068\u4ee5\u4e0b\u306e\u3088\u3046\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\r\n

\\[ W_i = \\begin{cases} \\frac{1}{P(T_i=1|X_i)} & \\text{if } T_i=1 \\text{ (\u4ecb\u5165\u7fa4)} \\\\ \\frac{1}{1 – P(T_i=1|X_i)} & \\text{if } T_i=0 \\text{ (\u7d71\u5236\u7fa4)} \\end{cases} \\]<\/p>\r\nBigQuery ML\u306e\u58c1\u3068\u300c\u7269\u7406\u7684\u30ea\u30b5\u30f3\u30d7\u30ea\u30f3\u30b0\u300d\u306b\u3088\u308b\u7a81\u7834<\/b>\r\n

\u5148\u8ff0\u306e\u901a\u308aBQML\u306b\u306f\u904b\u7528\u4e0a\u306e\u7d76\u5927\u306a\u30e1\u30ea\u30c3\u30c8\u304c\u3042\u308a\u307e\u3059\u304c\u3001\u4e00\u65b9\u3067\u6280\u8853\u7684\u306a\u58c1\u306b\u3082\u3076\u3064\u304b\u308a\u307e\u3059\u3002\u305d\u308c\u306f\u3001BQML\u306e\u30ed\u30b8\u30b9\u30c6\u30a3\u30c3\u30af\u56de\u5e30\uff08LOGISTIC_REG\uff09\u304c\u3001\u73fe\u6642\u70b9\u3067\u4efb\u610f\u306e\u91cd\u307f\u4ed8\u3051\u30ab\u30e9\u30e0\uff08Weight Column\uff09\u3092\u76f4\u63a5\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u3044\u306a\u3044\u70b9\u3067\u3059\u3002<\/p>\r\n

\u305d\u3053\u3067\u4eca\u56de\u306f\u3001\u300c\u7b97\u51fa\u3055\u308c\u305f\u91cd\u307f\u306e\u6570\u3060\u3051\u3001\u884c\u30c7\u30fc\u30bf\u3092\u7269\u7406\u7684\u306b\u8907\u88fd\uff08\u5897\u5e45\uff09\u3059\u308b\u300d\u3068\u3044\u3046\u529b\u6280\uff08\u7591\u4f3cIPW\uff09\u3092\u63a1\u7528\u3057\u307e\u3059\u3002\u5b9f\u969b\u306eSQL\u306e\u4e00\u90e8\u3092\u898b\u3066\u307f\u307e\u3057\u3087\u3046\u3002<\/p>\r\n

-- 1. \u50be\u5411\u30b9\u30b3\u30a2\uff08\u30af\u30fc\u30dd\u30f3\u914d\u5e03\u78ba\u7387\uff09\u306e\u7b97\u51fa\r\nCREATE OR REPLACE MODEL `ecommerce_project.ml_dataset.propensity_model` \r\nOPTIONS(model_type='LOGISTIC_REG', input_label_cols=['is_treated']) AS \r\nSELECT is_treated, customer_segment, ltv_percentile, past_orders_90d \/* ... *\/\r\nFROM `ecommerce_project.ml_dataset.mart_training_base`;\r\n\r\n-- 2. \u91cd\u307f\u306e\u8a08\u7b97\u3068\u884c\u306e\u7269\u7406\u7684\u8907\u88fd\uff08\u7591\u4f3cIPW\uff09\r\nCREATE OR REPLACE TABLE `ecommerce_project.ml_dataset.mart_weighted_data` AS\r\nWITH preds AS (\r\n  SELECT *, prob AS propensity_score \r\n  FROM ML.PREDICT(MODEL `ecommerce_project.ml_dataset.propensity_model`, ...)\r\n),\r\nweighted AS (\r\n  SELECT *,\r\n    CASE \r\n      WHEN is_treated = 1 THEN 1.0 \/ GREATEST(propensity_score, 0.01) \r\n      ELSE 1.0 \/ GREATEST(1.0 - propensity_score, 0.01) \r\n    END AS ipw_weight \r\n  FROM preds\r\n)\r\n-- \u91cd\u307f\u3092\u6574\u6570\u306b\u4e38\u3081\uff08\u6700\u592710\u500d\u306b\u5236\u9650\uff09\u3001GENERATE_ARRAY\u3067\u7269\u7406\u7684\u306b\u884c\u3092\u5897\u5e45\r\nSELECT * EXCEPT(propensity_score, ipw_weight)\r\nFROM (\r\n  SELECT *, CAST(LEAST(ROUND(ipw_weight), 10) AS INT64) AS rep_count\r\n  FROM weighted\r\n), UNNEST(GENERATE_ARRAY(1, GREATEST(rep_count, 1))) AS replica_id;<\/code><\/pre>\r\n
\u3053\u306e\u5897\u5e45\u3055\u308c\u305f\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u4f7f\u3063\u3066\u3001\u4ecb\u5165\u7fa4\uff08Treated\uff09\u3068\u7d71\u5236\u7fa4\uff08Control\uff09\u305d\u308c\u305e\u308c\u306e\u4e88\u6e2c\u30e2\u30c7\u30eb\u3092\u5b66\u7fd2\u3055\u305b\u307e\u3059\u3002<\/p>\n\n
Qini Curve\u306b\u3088\u308b\u30e2\u30c7\u30eb\u8a55\u4fa1<\/h2>\n
\u4f5c\u6210\u3057\u305f\u30e2\u30c7\u30eb\u304c\u672c\u5f53\u306b\u300c\u52b9\u679c\u306e\u9ad8\u3044\u9867\u5ba2\u300d\u3092\u898b\u3064\u3051\u51fa\u305b\u3066\u3044\u308b\u306e\u304b\uff1f\u305d\u308c\u3092\u8996\u899a\u7684\u306b\u8a55\u4fa1\u3059\u308b\u306e\u304c Qini Curve<\/span> \u3067\u3059\u3002<\/p>\r\n
\"\"<\/p>\r\n
\u30fb\u6a2a\u8ef8<\/strong>: Uplift Score\uff08\u4ecb\u5165\u7fa4\u306e\u4e88\u6e2c\u78ba\u7387 – \u7d71\u5236\u7fa4\u306e\u4e88\u6e2c\u78ba\u7387\uff09\u304c\u9ad8\u3044\u9806\u306b\u4e26\u3079\u305f\u9867\u5ba2\u306e\u30d1\u30fc\u30bb\u30f3\u30bf\u30a4\u30eb<\/p>\r\n
\u30fb\u7e26\u8ef8<\/strong>: \u7d2f\u7a4d\u306eUplift\uff08\u7d14\u5897\u52b9\u679c\uff09<\/p>\r\n
\u9752\u3044\u7dda\uff08\u30e2\u30c7\u30eb\u306e\u4e88\u6e2c\uff09\u304c\u6c34\u8272\u306e\u7dda\uff08\u30e9\u30f3\u30c0\u30e0\u914d\u5e03\uff09\u3092\u5927\u304d\u304f\u4e0a\u56de\u308a\u3001\u5de6\u4e0a\u306b\u81a8\u3089\u3093\u3067\u3044\u308b<\/strong>\u3053\u3068\u304c\u308f\u304b\u308a\u307e\u3059\u3002\u3053\u308c\u306f\u3001\u30e2\u30c7\u30eb\u304c\u300c\u30af\u30fc\u30dd\u30f3\u306b\u53cd\u5fdc\u3057\u3084\u3059\u3044\u30e6\u30fc\u30b6\u30fc\u300d\u3092\u6b63\u78ba\u306b\u4e0a\u4f4d\u306b\u30e9\u30f3\u30af\u4ed8\u3051\u3067\u304d\u3066\u3044\u308b\u8a3c\u62e0\u3067\u3059\u3002<\/p>\r\n
<\/p>\n\n
ROI\u30b7\u30df\u30e5\u30ec\u30fc\u30b7\u30e7\u30f3\uff1a\u7d50\u5c40\u3001\u3069\u3053\u307e\u3067\u914d\u308b\u306e\u304c\u6b63\u89e3\u304b\uff1f<\/h2>\n
\u7cbe\u5ea6\u306e\u9ad8\u3044\u30e2\u30c7\u30eb\u304c\u3067\u304d\u3066\u3082\u3001\u30d3\u30b8\u30cd\u30b9\u4e0a\u306e\u6700\u5927\u306e\u95a2\u5fc3\u4e8b\u306f\u300c\u5229\u76ca\u300d\u3067\u3059\u3002<\/p>\r\n
\u30af\u30fc\u30dd\u30f3\u914d\u5e03\u306b\u306f\u539f\u4fa1\uff08\u5024\u5f15\u304d\u30b3\u30b9\u30c8\uff09\u304c\u304b\u304b\u308a\u307e\u3059\u3002\u305d\u306e\u305f\u3081\u3001\u5404\u30e6\u30fc\u30b6\u30fc\u3078\u306e\u671f\u5f85\u5229\u76ca\uff08Expected Net Value\uff09\u3092\u4ee5\u4e0b\u306e\u6570\u5f0f\u3067\u5b9a\u7fa9\u3057\u307e\u3059\u3002<\/p>\r\n
$$ \\begin{aligned} Expected\\_Net\\_Value &= LTV \\times \\Big( Uplift\\_Score – P(CV|Treated) \\times Coupon\\_Cost\\_Rate \\Big) \\end{aligned} $$<\/p>\r\n
\u3053\u306e\u671f\u5f85\u5229\u76ca\u3092\u30d9\u30fc\u30b9\u306b\u3001\u4e0a\u4f4d\u304b\u3089\u9806\u306b\u30af\u30fc\u30dd\u30f3\u3092\u914d\u3063\u305f\u5834\u5408\u306e\u30b7\u30df\u30e5\u30ec\u30fc\u30b7\u30e7\u30f3\u3092\u884c\u3063\u305f\u306e\u304c\u4ee5\u4e0b\u306e\u30b0\u30e9\u30d5\u3067\u3059\u3002<\/p>\r\n
\"\"<\/p>\r\n
    \r\n\t
  • \u9752\u8272\u306e\u6298\u308c\u7dda\uff08Cumulative Profit\uff09<\/strong>: \u7d2f\u7a4d\u306e\u671f\u5f85\u5229\u76ca<\/li>\r\n\t
  • \u6c34\u8272\u306e\u68d2\u30b0\u30e9\u30d5\uff08Marginal Profit\uff09<\/strong>: \u9650\u754c\u5229\u76ca\uff08\u30e6\u30fc\u30b6\u30fc1\u4eba\u8ffd\u52a0\u3054\u3068\u306e\u5229\u76ca\u5897\u6e1b\uff09<\/li>\r\n<\/ul>\r\n
    \u30b0\u30e9\u30d5\u3092\u898b\u308b\u3068\u3001\u4e0a\u4f4d\u304b\u3089\u30af\u30fc\u30dd\u30f3\u3092\u914d\u5e03\u3057\u3066\u3044\u304f\u3068\u7d2f\u7a4d\u5229\u76ca\uff08\u9752\u7dda\uff09\u306f\u4e0a\u6607\u3092\u7d9a\u3051\u307e\u3059\u304c\u300190%\u4ed8\u8fd1\u3092\u30d4\u30fc\u30af\uff08\u6975\u5927\u5024\uff09\u306b\u4e0b\u964d\u306b\u8ee2\u3058\u3066\u3044\u307e\u3059\u3002\u53f3\u7aef\u306e\u5c64\uff08\u68d2\u30b0\u30e9\u30d5\u304c\u30de\u30a4\u30ca\u30b9\u306b\u6c88\u3093\u3067\u3044\u308b\u5c64\uff09\u306f\u3001\u300c\u30af\u30fc\u30dd\u30f3\u3092\u914d\u308b\u3068\u9006\u306b\u5229\u76ca\u304c\u8d64\u5b57\u306b\u306a\u308b\u5c64\u300d\u3067\u3059\u3002<\/p>\r\n
    \u3010\u7d50\u8ad6\uff08\u30a2\u30af\u30b7\u30e7\u30f3\u30d7\u30e9\u30f3\uff09\u3011<\/strong>
    \r\n\u3053\u306e\u30b7\u30df\u30e5\u30ec\u30fc\u30b7\u30e7\u30f3\u7d50\u679c\u306b\u3088\u308a\u3001\u300cUplift Score\u4e0a\u4f4d90%\u306e\u30e6\u30fc\u30b6\u30fc\u3092\u30bf\u30fc\u30b2\u30c3\u30c8\u3068\u3059\u308b\u3053\u3068\u3067\u3001ROI\u3092\u6700\u5927\u5316\u3067\u304d\u308b\u300d\u3068\u3044\u3046\u660e\u78ba\u306a\u610f\u601d\u6c7a\u5b9a\u304c\u53ef\u80fd\u306b\u306a\u308a\u307e\u3059\u3002\u6b8b\u308a\u306e10%\uff08\u30de\u30a4\u30ca\u30b9\u5c64\uff09\u306f\u914d\u5e03\u5bfe\u8c61\u5916\uff08\u63a8\u5968\u30d5\u30e9\u30b0\uff1d0\uff09\u3068\u3057\u3066\u9664\u5916\u3057\u307e\u3059\u3002<\/p>\n\n
    \u307e\u3068\u3081<\/h2>\n
    \u6a5f\u68b0\u5b66\u7fd2\u306b\u3088\u308b\u4e88\u6e2c\u306f\u3001\u78ba\u7387\u3084\u7cbe\u5ea6\u3092\u51fa\u3059\u307e\u3067\u304c\u4ed5\u4e8b\u3067\u306f\u3042\u308a\u307e\u305b\u3093\u3002\u305d\u308c\u3092\u300c\u30d3\u30b8\u30cd\u30b9\u306e\u5229\u76ca\uff08ROI\uff09\u3068\u30a2\u30af\u30b7\u30e7\u30f3\u300d\u306b\u5909\u63db\u3059\u308b\u3053\u3068\u3067\u521d\u3081\u3066\u4fa1\u5024\u3092\u751f\u307f\u307e\u3059\u3002<\/p>\r\n
    \uff11\uff0e\u30c7\u30fc\u30bf\u30a6\u30a7\u30a2\u30cf\u30a6\u30b9\u5185\u3067\u5b8c\u7d50\u3059\u308bBQML\u306e\u30a8\u30b3\u30b7\u30b9\u30c6\u30e0<\/strong>\u3092\u6d3b\u7528\u3057\u3001\u904b\u7528\u30b3\u30b9\u30c8\u3092\u5287\u7684\u306b\u4e0b\u3052\u308b\u3002<\/p>\r\n
    \uff12\uff0eQini Curve<\/strong>\u3067\u30e2\u30c7\u30eb\u306e\u6027\u80fd\u3092\u5ba2\u89b3\u7684\u306b\u8a55\u4fa1\u3059\u308b\u3002<\/p>\r\n
    \uff13\uff0eROI\u30b7\u30df\u30e5\u30ec\u30fc\u30b7\u30e7\u30f3<\/strong>\u306b\u3088\u3063\u3066\u3001\u5229\u76ca\u304c\u6700\u5927\u5316\u3059\u308b\u300c\u640d\u76ca\u5206\u5c90\u70b9\u300d\u3092\u7279\u5b9a\u3057\u3001\u81ea\u52d5\u914d\u4fe1\u306e\u30bb\u30b0\u30e1\u30f3\u30c8\u3092\u5207\u308b\u3002<\/p>\r\n
    \u30c7\u30fc\u30bf\u5206\u6790\u30c1\u30fc\u30e0\u304c\u300c\u3068\u308a\u3042\u3048\u305aA\/B\u30c6\u30b9\u30c8\u300d\u3068\u3044\u3046\u601d\u8003\u304b\u3089\u8131\u5374\u3057\u3001\u3053\u306e\u3088\u3046\u306a\u30aa\u30d5\u30e9\u30a4\u30f3\u691c\u8a3c\u3068\u671f\u5f85\u5024\u30b7\u30df\u30e5\u30ec\u30fc\u30b7\u30e7\u30f3\u3092\u63d0\u793a\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308c\u3070\u3001\u30de\u30fc\u30b1\u30c6\u30a3\u30f3\u30b0\u65bd\u7b56\u306e\u78ba\u5ea6\u3068\u30b9\u30d4\u30fc\u30c9\u306f\u5287\u7684\u306b\u5411\u4e0a\u3059\u308b\u306f\u305a\u3067\u3059\u3002<\/p>\r\n
    <\/p>","protected":false},"excerpt":{"rendered":"
    \u3080\u3084\u307f\u306a\u30af\u30fc\u30dd\u30f3\u914d\u5e03\u3067\u5229\u76ca\u3092\u524a\u3063\u3066\u3044\u307e\u305b\u3093\u304b\uff1f\u56e0\u679c\u63a8\u8ad6\u300cUplift Modeling\u300d\u3092\u7528\u3044\u3066\u300c\u30af\u30fc\u30dd\u30f3\u3092\u914d\u308b\u3079\u304d\u9867\u5ba2\u300d\u3060\u3051\u3092\u7279\u5b9a\u3057\u3001\u671f\u5f85\u5229\u76ca(ROI)\u3092\u6700\u5927\u5316\u3059\u308b\u624b\u6cd5\u3092\u5b9f\u8df5\u89e3\u8aac\u3057\u307e\u3059\u3002BigQuery ML\u306e\u5236\u7d04\u3092\u30cf […]<\/p>\n","protected":false},"author":89,"featured_media":5665,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"content-type":"","swell_btn_cv_data":"","footnotes":"","_wp_rev_ctl_limit":""},"categories":[1246],"tags":[484,33,419],"class_list":["post-7730","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-data-infrastructure","tag-484","tag-33","tag-419"],"_links":{"self":[{"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/posts\/7730","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/users\/89"}],"replies":[{"embeddable":true,"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/comments?post=7730"}],"version-history":[{"count":0,"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/posts\/7730\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/media\/5665"}],"wp:attachment":[{"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/media?parent=7730"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/categories?post=7730"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/tags?post=7730"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}