feat:V566实验按供给类型+驱动策略对item加权

基于V563,从alg_vid_feature_basic_info读取supply_type/driving_strategy,硬编码3条规则:
- UGC/垂直spider 降权0.8
- 自动AGC+当下供需gap 提权1.5
- 人工AGC+(人工历史需求|当下供需gap) 提权1.2
3个字段值打入scoresMap用于埋点观测。

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

recommend-server-service/src/main/java/com/tzld/piaoquan/recommend/server/service/rank/strategy/RankStrategy4RegionMergeModelV566.java

@@ -29,6 +29,15 @@ import java.util.stream.Stream;
 
 @Service
 @Slf4j
+/**
+ * V566 实验：基于 V563,按"供给类型(supply_type) + 驱动策略(driving_strategy)"对 item 加权/降权。
+ * 数据来源: alg_vid_feature_basic_info.feature JSON。
+ * 规则(硬编码,见 {@link #getSupplyWeight}):
+ *   - supply_type ∈ {UGC, 垂直spider}                          → 0.8
+ *   - supply_type = 自动AGC AND driving_strategy = 当下供需gap → 1.5
+ *   - supply_type = 人工AGC AND driving_strategy ∈ {人工历史需求, 当下供需gap} → 1.2
+ *   - 其他                                                       → 1.0
+ */
 public class RankStrategy4RegionMergeModelV566 extends RankStrategy4RegionMergeModelBasic {
     @ApolloJsonValue("${rank.score.merge.weightv566:}")
     private Map<String, Double> mergeWeight;
@@ -36,6 +45,10 @@ public class RankStrategy4RegionMergeModelV566 extends RankStrategy4RegionMergeM
     @Autowired
     private FeatureService featureService;
 
+    // V566 供给加权规则用到的常量
+    private static final Set<String> SUPPLY_DEMOTE_TYPES = new HashSet<>(Arrays.asList("UGC", "垂直spider"));
+    private static final Set<String> AGC_BOOST_DRIVINGS = new HashSet<>(Arrays.asList("人工历史需求", "当下供需gap"));
+
     @Override
     public List<Video> mergeAndRankRovRecall(RankParam param) {
         Map<String, Double> mergeWeight = this.mergeWeight != null ? this.mergeWeight : new HashMap<>(0);
@@ -136,7 +149,7 @@ public class RankStrategy4RegionMergeModelV566 extends RankStrategy4RegionMergeM
 
         // 4. 排序模型计算
         Map<String, Float> sceneFeatureMap = new HashMap<>(0);
-        List<RankItem> items = ScorerUtils.getScorerPipeline("feeds_score_config_str_ros_dnn_20260424.conf").scoring(sceneFeatureMap, userFeatureMap, rankItems);
+        List<RankItem> items = ScorerUtils.getScorerPipeline("feeds_score_config_dnn_20260407.conf").scoring(sceneFeatureMap, userFeatureMap, rankItems);
 
         // 5. 排序公式特征
         double xgbRovNegRate = mergeWeight.getOrDefault("xgbRovNegRate", 0.059);
@@ -149,9 +162,6 @@ public class RankStrategy4RegionMergeModelV566 extends RankStrategy4RegionMergeM
         double vorAdd = mergeWeight.getOrDefault("vor_add", 0.1d);
         double vorW = mergeWeight.getOrDefault("vor_w", 1.0d);
 
-        double leaveW = mergeWeight.getOrDefault("leave_w", 1d);
-        double leaveExp = mergeWeight.getOrDefault("leave_exp", 1d);
-
         double c1Rovn1hW = mergeWeight.getOrDefault("c1_rovn_1h_w", 0d);
         double c1Rovn24hW = mergeWeight.getOrDefault("c1_rovn_24h_w", 0d);
 
@@ -183,8 +193,9 @@ public class RankStrategy4RegionMergeModelV566 extends RankStrategy4RegionMergeM
             double hasReturnRovScore = Double.parseDouble(vid2MapFeature.getOrDefault(item.getVideoId() + "", new HashMap<>()).getOrDefault("rov", "0"));
             item.getScoresMap().put("hasReturnRovScore", hasReturnRovScore);
 
-            double norXGBScore = item.getScoresMap().getOrDefault("NorXGBScore", 0d);
-            double newNorXGBScore = norPowerCalibration(xgbNorPowerWeight, xgbNorPowerExp, norXGBScore);
+            double norDNNScore = item.getScoresMap().getOrDefault("NorDNNScore", 0d);
+            double newNorDNNScore = norPowerCalibration(xgbNorPowerWeight, xgbNorPowerExp, norDNNScore);
+            item.getScoresMap().put("newNorDNNScore", newNorDNNScore);
             item.getScoresMap().put("rosAdd", rosAdd);
             item.getScoresMap().put("rosW", rosW);
 
@@ -193,12 +204,6 @@ public class RankStrategy4RegionMergeModelV566 extends RankStrategy4RegionMergeM
             item.getScoresMap().put("vorAdd", vorAdd);
             item.getScoresMap().put("vorW", vorW);
 
-            double pLeave = item.getScoresMap().getOrDefault("pLeave", 0d);
-            double newPLeave = Math.pow((1 - leaveW * pLeave), leaveExp);
-            item.getScoresMap().put("leaveW", leaveW);
-            item.getScoresMap().put("leaveExp", leaveExp);
-            item.getScoresMap().put("newPLeave", newPLeave);
-
             Map<String, String> bcData = videoBCData.getOrDefault(String.valueOf(item.getVideoId()), new HashMap<>()).getOrDefault("alg_vid_feature_b_c_data", new HashMap<>());
             Map<String, String> cdNData = videoBCData.getOrDefault(String.valueOf(item.getVideoId()), new HashMap<>()).getOrDefault("alg_vid_feature_cn_dn_data", new HashMap<>());
 
@@ -248,9 +253,26 @@ public class RankStrategy4RegionMergeModelV566 extends RankStrategy4RegionMergeM
             item.getScoresMap().put("dnRovn24hW", dnRovn24hW);
             item.getScoresMap().put("dnRovn24h", dnRovn24h);
 
-            score = fmRov * (rosAdd + rosW * newNorXGBScore) * (vorAdd + vorW * vor) * newPLeave + c1RovnScore + b0StrScore + b0RorScore + cnRovnScore + dnRovnScore;
+            score = fmRov * (rosAdd + rosW * newNorDNNScore) * (vorAdd + vorW * vor) + c1RovnScore + b0StrScore + b0RorScore + cnRovnScore + dnRovnScore;
 
+            // ============== V566: 按供给类型 + 驱动策略加权 ==============
+            // 数据来源: alg_vid_feature_basic_info.feature.{supply_type, driving_strategy}
+            // 规则硬编码在 getSupplyWeight() 中
             Video video = item.getVideo();
+            Map<String, String> vidBaseInfo = videoBaseInfoMap
+                    .getOrDefault(String.valueOf(item.getVideoId()), Collections.emptyMap())
+                    .getOrDefault("alg_vid_feature_basic_info", Collections.emptyMap());
+            String supplyType = vidBaseInfo.getOrDefault("supply_type", "unknown");
+            String drivingStrategy = vidBaseInfo.getOrDefault("driving_strategy", "unknown");
+            String sceneDimension = vidBaseInfo.getOrDefault("scene_dimension", "unknown");
+            double supplyTypeWeightValue = getSupplyWeight(supplyType, drivingStrategy);
+            item.getScoresMap().put("supplyType_" + supplyType, 1.0);
+            item.getScoresMap().put("drivingStrategy_" + drivingStrategy, 1.0);
+            item.getScoresMap().put("sceneDimension_" + sceneDimension, 1.0);
+            item.getScoresMap().put("supplyTypeWeight", supplyTypeWeightValue);
+            score = score * supplyTypeWeightValue;
+            // ============== V566 加权块结束 ==============
+
             video.setScore(score);
             video.setSortScore(score);
             video.setScoresMap(item.getScoresMap());
@@ -382,7 +404,6 @@ public class RankStrategy4RegionMergeModelV566 extends RankStrategy4RegionMergeM
                     Map<String, Float> videoFeature = getVideoFeature(currentMs, vid, userProfile, creativeInfo, headInfo, rankInfo, c7Map, c8Map, userOriginInfo, historyVideoMap, videoOriginInfo);
                     featureMap.putAll(videoFeature);
                     item.featureMap = featureMap;
-                    item.norFeatureMap = item.featureMap;
 
                     Map<String, String> userNetworkSeqFeature = userOriginInfo.getOrDefault("alg_user_network_seq_feature", new HashMap<>());
 
@@ -463,4 +484,21 @@ public class RankStrategy4RegionMergeModelV566 extends RankStrategy4RegionMergeM
         }
         return newScore;
     }
+
+    /**
+     * V566 供给加权规则(硬编码)。
+     * 规则顺序固定:命中即返回,按业务优先级从高到低。
+     */
+    private double getSupplyWeight(String supplyType, String drivingStrategy) {
+        if (SUPPLY_DEMOTE_TYPES.contains(supplyType)) {
+            return 0.8;
+        }
+        if ("自动AGC".equals(supplyType) && "当下供需gap".equals(drivingStrategy)) {
+            return 1.5;
+        }
+        if ("人工AGC".equals(supplyType) && AGC_BOOST_DRIVINGS.contains(drivingStrategy)) {
+            return 1.2;
+        }
+        return 1.0;
+    }
 }