增强ros-统计量影响

recommend-server-service/src/main/java/com/tzld/piaoquan/recommend/server/service/rank/strategy/RankStrategy4RegionMergeModelV565.java

@@ -109,12 +109,13 @@ public class RankStrategy4RegionMergeModelV565 extends RankStrategy4RegionMergeM
 
         // 4. 排序模型计算
         Map<String, Float> sceneFeatureMap = new HashMap<>(0);
-        List<RankItem> items = ScorerUtils.getScorerPipeline("feeds_score_config_fm_xgb_20250729.conf").scoring(sceneFeatureMap, userFeatureMap, userFeatureMap, rankItems);
+        List<RankItem> items = ScorerUtils.getScorerPipeline("feeds_score_config_fm_xgb_20260116.conf").scoring(sceneFeatureMap, userFeatureMap, userFeatureMap, rankItems);
 
         // 5. 排序公式特征
         double xgbRovNegRate = mergeWeight.getOrDefault("xgbRovNegRate", 0.059);
         double xgbNorPowerWeight = mergeWeight.getOrDefault("xgbNorPowerWeight", 1.22);
         double xgbNorPowerExp = mergeWeight.getOrDefault("xgbNorPowerExp", 1.15);
+        double rosMinusAlpha = mergeWeight.getOrDefault("rosMinusAlpha", 1.1);
         Map<String, Map<String, String>> vid2MapFeature = this.getVideoRedisFeature(vids, "redis:vid_hasreturn_vor:");
 
         // 获取权重
@@ -132,29 +133,25 @@ public class RankStrategy4RegionMergeModelV565 extends RankStrategy4RegionMergeM
 
         List<Video> result = new ArrayList<>();
         for (RankItem item : items) {
-            String vidMergeCate2 = this.findVideoMergeCate2(videoBaseInfoMap, String.valueOf(item.getVideoId()));
-            double strPlusCalibrationCoefficient = this.calcStrPlusCalibrationCoefficientByMergeCate2(vidMergeCate2);
-
             double score;
             double fmRovOrigin = item.getScoreRov();
             item.getScoresMap().put("fmRovOrigin", fmRovOrigin);
-            double fmRovCoefficientBefore = restoreScore(fmRovOrigin, xgbRovNegRate);
-            item.getScoresMap().put("fmRovCoefficientBefore", fmRovCoefficientBefore);
-            double fmRov = fmRovCoefficientBefore * strPlusCalibrationCoefficient;
+            double fmRov = restoreScore(fmRovOrigin, xgbRovNegRate);
             item.getScoresMap().put("fmRov", fmRov);
-            item.getScoresMap().put("strPlusCalibrationCoefficient", strPlusCalibrationCoefficient);
             double hasReturnRovScore = Double.parseDouble(vid2MapFeature.getOrDefault(item.getVideoId() + "", new HashMap<>()).getOrDefault("rov", "0"));
+            double newRosMinusScore = scalaRosMinusScore(hasReturnRovScore, rosMinusAlpha);
             item.getScoresMap().put("hasReturnRovScore", hasReturnRovScore);
             double norXGBScore = item.getScoresMap().getOrDefault("NorXGBScore", 0d);
             double newNorXGBScore = norPowerCalibration(xgbNorPowerWeight, xgbNorPowerExp, norXGBScore);
             double vor = Double.parseDouble(vid2MapFeature.getOrDefault(item.getVideoId() + "", new HashMap<>()).getOrDefault("vor", "0"));
             item.getScoresMap().put("vor", vor);
 
+            String vidMergeCate2 = this.findVideoMergeCate2(videoBaseInfoMap, String.valueOf(item.getVideoId()));
             Double scoreCoefficient = cate2Coefficient.getOrDefault(vidMergeCate2, 0d);
             item.getScoresMap().put("scoreCoefficient", scoreCoefficient);
             item.getScoresMap().put("cate2CoefficientDenominator", cate2CoefficientDenominator);
 
-            score = fmRov * (0.1 + newNorXGBScore) * (0.1 + vor) * (1 + scoreCoefficient / cate2CoefficientDenominator);
+            score = fmRov * (0.1 + newRosMinusScore) * (0.1 + vor) * (1 + scoreCoefficient / cate2CoefficientDenominator);
 
             Video video = item.getVideo();
             video.setScore(score);
@@ -342,6 +339,13 @@ public class RankStrategy4RegionMergeModelV565 extends RankStrategy4RegionMergeM
         return baseInfo;
     }
 
+    private double scalaRosMinusScore(double rosMinusScore, double alpha) {
+        if (rosMinusScore > 0) {
+            return Math.pow(rosMinusScore, alpha);
+        }
+        return rosMinusScore;
+    }
+
     private double norPowerCalibration(double weight, double exp, double score) {
         double newScore = weight * Math.pow(score, exp);
         if (newScore > 100) {