File size: 2,226 Bytes
6f159f9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
{
  "model": "LightGBM LambdaRank V3",
  "description": "Trained on cross-survey authority labels with 13 graph-based features",
  "num_features": 13,
  "features": [
    "candidate_num_cited_by",
    "candidate_num_references",
    "survey_num_references",
    "co_citation_count",
    "bibliographic_coupling",
    "jaccard_refs",
    "adamic_adar",
    "n_surveys_citing",
    "n_total_citations",
    "cited_by_survey_refs",
    "candidate_cites_survey_refs",
    "shared_citers_with_survey",
    "hub_score"
  ],
  "training": {
    "train_rows": 58749,
    "train_queries": 486,
    "eval_rows": 1825,
    "eval_queries": 14,
    "best_iteration": 1,
    "training_time_seconds": 0.4
  },
  "eval_metrics": {
    "ndcg@10": 0.9515,
    "ndcg@20": 0.9607,
    "mrr": 0.9524,
    "hard_neg_auc": 0.7364
  },
  "train_metrics": {
    "ndcg@10": 0.9494,
    "hard_neg_auc": 0.738
  },
  "comparison": {
    "Random": {
      "ndcg@10": 0.2356,
      "hard_neg_auc": 0.5039
    },
    "CitationCount": {
      "ndcg@10": 0.8371,
      "hard_neg_auc": 0.6277
    },
    "Authority": {
      "ndcg@10": 0.929,
      "hard_neg_auc": 0.6463
    },
    "V3_LightGBM": {
      "ndcg@10": 0.9494,
      "hard_neg_auc": 0.738
    },
    "Oracle": {
      "ndcg@10": 1.0,
      "hard_neg_auc": 1.0
    }
  },
  "feature_importance": [
    {
      "feature": "n_surveys_citing",
      "importance": 1607.3
    },
    {
      "feature": "hub_score",
      "importance": 39.8
    },
    {
      "feature": "survey_num_references",
      "importance": 33.0
    },
    {
      "feature": "adamic_adar",
      "importance": 5.8
    },
    {
      "feature": "candidate_num_references",
      "importance": 3.3
    },
    {
      "feature": "jaccard_refs",
      "importance": 2.7
    },
    {
      "feature": "co_citation_count",
      "importance": 2.4
    }
  ],
  "key_insight": "V3 improves hard_neg_auc from 0.628 (CitationCount) to 0.738 (+17.5% relative). This means V3 can tell the difference between papers an expert included vs excluded 73.8% of the time, vs 62.8% for pure popularity. The n_surveys_citing feature dominates \u2014 cross-survey consensus is the strongest signal for paper importance beyond raw citation count."
}