quantumly commited on
Commit
c480fbb
·
verified ·
1 Parent(s): d8e37b3

v0.2 appraiser: trained 2026-04-26

Browse files
Files changed (1) hide show
  1. v0_metadata.json +319 -0
v0_metadata.json ADDED
@@ -0,0 +1,319 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trained_at": "2026-04-26T06:46:09.798149+00:00",
3
+ "data_run_date": "2026-04-25",
4
+ "version": "v0.2-leakage-fixed",
5
+ "fixes_vs_v0_1": [
6
+ "prior_transfer_count replaces lifetime_transfer_count (no future leakage)",
7
+ "$10 price floor (was $1) drops wash-trade noise",
8
+ "fixed-window temporal split (train < 2023-10-01, val Q4 2023, test 2024+)",
9
+ "recency-weighted sample weights (1y half-life)"
10
+ ],
11
+ "splits": {
12
+ "train": {
13
+ "rows": 265240,
14
+ "start": "2022-01-28",
15
+ "end": "2023-09-30"
16
+ },
17
+ "val": {
18
+ "rows": 3545,
19
+ "start": "2023-10-01",
20
+ "end": "2023-12-31"
21
+ },
22
+ "test": {
23
+ "rows": 2744,
24
+ "start": "2024-01-01",
25
+ "end": "2024-05-04"
26
+ }
27
+ },
28
+ "feature_count": 146,
29
+ "feature_cols": [
30
+ "len",
31
+ "n_digits",
32
+ "n_letters",
33
+ "n_special",
34
+ "n_lower",
35
+ "n_upper",
36
+ "is_palindrome",
37
+ "is_all_digits",
38
+ "is_all_letters",
39
+ "is_ascii",
40
+ "has_unicode",
41
+ "starts_digit",
42
+ "ends_digit",
43
+ "max_char_run",
44
+ "n_unique_chars",
45
+ "in_wikipedia",
46
+ "in_geonames",
47
+ "in_us_firstname",
48
+ "in_iso3166",
49
+ "in_ticker",
50
+ "in_sec_edgar",
51
+ "in_wiktionary_en",
52
+ "wordlist_hits",
53
+ "club__prepunk_full_rankings",
54
+ "club__personas",
55
+ "club__gamertags",
56
+ "club__luxury",
57
+ "club__top_nouns",
58
+ "club__familynames_usa",
59
+ "club__wikidata_top_fantasy_char",
60
+ "club__firstnames_usa",
61
+ "club__natural_wonders",
62
+ "club__logistics",
63
+ "club__social_handles",
64
+ "club__top500_cities_usa",
65
+ "club__common_english",
66
+ "club__finance_terms",
67
+ "club__performing_arts",
68
+ "club__paranormal",
69
+ "club__gamertags_double",
70
+ "club__memes",
71
+ "club__sports",
72
+ "club__home",
73
+ "club__top_crypto_names",
74
+ "club__top500_cities_global",
75
+ "club__catholicism",
76
+ "club__country_codes",
77
+ "club__fine_art",
78
+ "club__us_government",
79
+ "club__pokemon_gen3",
80
+ "club__pokemon_gen4",
81
+ "club__crypto_terms",
82
+ "club__top_crypto_tickers",
83
+ "club__crayola_classic",
84
+ "club__us_states",
85
+ "club__mythical_creatures",
86
+ "club__historic_figures",
87
+ "club__pokemon_gen2",
88
+ "club__holidays",
89
+ "club__common_animals",
90
+ "club__pokemon_gen1",
91
+ "club__conspiracy_theories",
92
+ "club__gen_alpha",
93
+ "club__currency_symbols",
94
+ "club__currency_names",
95
+ "n_clubs",
96
+ "trademark_conflict",
97
+ "name_age_days",
98
+ "prior_transfer_count",
99
+ "fg_value",
100
+ "eth_tvl_usd",
101
+ "eth_stable_mcap",
102
+ "eth_dex_volume",
103
+ "nft_total_fee_usd",
104
+ "pca_000",
105
+ "pca_001",
106
+ "pca_002",
107
+ "pca_003",
108
+ "pca_004",
109
+ "pca_005",
110
+ "pca_006",
111
+ "pca_007",
112
+ "pca_008",
113
+ "pca_009",
114
+ "pca_010",
115
+ "pca_011",
116
+ "pca_012",
117
+ "pca_013",
118
+ "pca_014",
119
+ "pca_015",
120
+ "pca_016",
121
+ "pca_017",
122
+ "pca_018",
123
+ "pca_019",
124
+ "pca_020",
125
+ "pca_021",
126
+ "pca_022",
127
+ "pca_023",
128
+ "pca_024",
129
+ "pca_025",
130
+ "pca_026",
131
+ "pca_027",
132
+ "pca_028",
133
+ "pca_029",
134
+ "pca_030",
135
+ "pca_031",
136
+ "pca_032",
137
+ "pca_033",
138
+ "pca_034",
139
+ "pca_035",
140
+ "pca_036",
141
+ "pca_037",
142
+ "pca_038",
143
+ "pca_039",
144
+ "pca_040",
145
+ "pca_041",
146
+ "pca_042",
147
+ "pca_043",
148
+ "pca_044",
149
+ "pca_045",
150
+ "pca_046",
151
+ "pca_047",
152
+ "pca_048",
153
+ "pca_049",
154
+ "pca_050",
155
+ "pca_051",
156
+ "pca_052",
157
+ "pca_053",
158
+ "pca_054",
159
+ "pca_055",
160
+ "pca_056",
161
+ "pca_057",
162
+ "pca_058",
163
+ "pca_059",
164
+ "pca_060",
165
+ "pca_061",
166
+ "pca_062",
167
+ "pca_063",
168
+ "knn_count",
169
+ "knn_mean_log",
170
+ "knn_median_log",
171
+ "knn_p90_log",
172
+ "knn_max_sim",
173
+ "knn_min_sim",
174
+ "knn_log_max",
175
+ "knn_log_min"
176
+ ],
177
+ "pca_dim": 64,
178
+ "embedding_model": "sentence-transformers/all-mpnet-base-v2",
179
+ "best_iteration": 170,
180
+ "xgb_params": {
181
+ "objective": "reg:squarederror",
182
+ "eval_metric": "rmse",
183
+ "tree_method": "hist",
184
+ "device": "cuda",
185
+ "max_depth": 7,
186
+ "learning_rate": 0.04,
187
+ "subsample": 0.85,
188
+ "colsample_bytree": 0.65,
189
+ "min_child_weight": 8,
190
+ "reg_alpha": 0.5,
191
+ "reg_lambda": 2.0,
192
+ "seed": 42
193
+ },
194
+ "metrics": {
195
+ "train": {
196
+ "r2_log": 0.7700075507164001,
197
+ "rmse_log": 0.7744038701057434,
198
+ "mae_log": 0.5210891962051392,
199
+ "median_ape": 0.32541945576667786,
200
+ "bias_log": -0.0004094731993973255
201
+ },
202
+ "val": {
203
+ "r2_log": 0.6602156162261963,
204
+ "rmse_log": 1.0678441524505615,
205
+ "mae_log": 0.7699634432792664,
206
+ "median_ape": 0.5703179836273193,
207
+ "bias_log": 0.2033758908510208
208
+ },
209
+ "test": {
210
+ "r2_log": 0.3080594539642334,
211
+ "rmse_log": 1.5469372272491455,
212
+ "mae_log": 1.2711551189422607,
213
+ "median_ape": 1.3828195333480835,
214
+ "bias_log": 0.7319739460945129
215
+ }
216
+ },
217
+ "top_features": [
218
+ {
219
+ "name": "knn_mean_log",
220
+ "gain": 1713.8084716796875
221
+ },
222
+ {
223
+ "name": "knn_p90_log",
224
+ "gain": 1612.9539794921875
225
+ },
226
+ {
227
+ "name": "len",
228
+ "gain": 1364.2567138671875
229
+ },
230
+ {
231
+ "name": "in_wikipedia",
232
+ "gain": 1052.1461181640625
233
+ },
234
+ {
235
+ "name": "is_all_digits",
236
+ "gain": 943.7355346679688
237
+ },
238
+ {
239
+ "name": "knn_median_log",
240
+ "gain": 604.246337890625
241
+ },
242
+ {
243
+ "name": "n_digits",
244
+ "gain": 337.8750305175781
245
+ },
246
+ {
247
+ "name": "pca_000",
248
+ "gain": 289.33624267578125
249
+ },
250
+ {
251
+ "name": "n_clubs",
252
+ "gain": 281.8310546875
253
+ },
254
+ {
255
+ "name": "ends_digit",
256
+ "gain": 276.56658935546875
257
+ },
258
+ {
259
+ "name": "name_age_days",
260
+ "gain": 274.4316101074219
261
+ },
262
+ {
263
+ "name": "n_unique_chars",
264
+ "gain": 267.8934020996094
265
+ },
266
+ {
267
+ "name": "club__gamertags",
268
+ "gain": 267.10980224609375
269
+ },
270
+ {
271
+ "name": "starts_digit",
272
+ "gain": 259.8104248046875
273
+ },
274
+ {
275
+ "name": "n_special",
276
+ "gain": 258.7098693847656
277
+ },
278
+ {
279
+ "name": "pca_001",
280
+ "gain": 238.66867065429688
281
+ },
282
+ {
283
+ "name": "n_letters",
284
+ "gain": 217.4210662841797
285
+ },
286
+ {
287
+ "name": "knn_log_max",
288
+ "gain": 214.86541748046875
289
+ },
290
+ {
291
+ "name": "trademark_conflict",
292
+ "gain": 213.1521453857422
293
+ },
294
+ {
295
+ "name": "pca_034",
296
+ "gain": 211.79525756835938
297
+ },
298
+ {
299
+ "name": "is_ascii",
300
+ "gain": 211.76641845703125
301
+ },
302
+ {
303
+ "name": "eth_stable_mcap",
304
+ "gain": 204.28631591796875
305
+ },
306
+ {
307
+ "name": "has_unicode",
308
+ "gain": 198.2250518798828
309
+ },
310
+ {
311
+ "name": "wordlist_hits",
312
+ "gain": 196.23956298828125
313
+ },
314
+ {
315
+ "name": "n_lower",
316
+ "gain": 189.0078887939453
317
+ }
318
+ ]
319
+ }