通常用を72時間学習させてみましたが、納得のいくRMSEは得られなかったorz 最適RMSE 1.4113はここ最近としてはまあまあなものですが
| 1,108R | 1点 | 芝(572R) | ダート(493R) | 障害(43R) | 8頭以下(110R) | 9～12頭(302R) | 13頭以上(696R) | 多点 |
| 単勝 | 23.56% (80.47%) |
22.55% (77.80%) |
23.53% (79.55%) |
37.21% (126.51%) |
41.82% (92.91%) |
25.83% (79.17%) |
19.68% (79.07%) |
50.81% (71.34%) |
| 複勝 | 51.53% (84.20%) |
50.87% (87.92%) |
51.12% (79.27%) |
65.12% (91.16%) |
69.09% (87.82%) |
59.60% (91.29%) |
45.26% (80.55%) |
85.20% (81.95%) |
| 枠連 | 12.98% (92.70%) |
10.62% (95.46%) |
14.00% (86.39%) |
32.26% (145.48%) |
-- (--) |
16.23% (69.24%) |
10.49% (95.16%) |
25.32% (75.99%) |
| 馬連 | 9.48% (58.35%) |
8.22% (49.90%) |
9.94% (61.38%) |
20.93% (136.05%) |
21.82% (74.09%) |
11.92% (61.16%) |
6.47% (54.64%) |
21.93% (63.38%) |
| ワイド | 20.04% (80.78%) |
18.53% (80.02%) |
20.69% (78.90%) |
32.56% (112.33%) |
42.73% (79.45%) |
24.83% (75.43%) |
14.37% (83.30%) |
41.52% (81.54%) |
| 馬単 | 5.78% (66.03%) |
4.55% (58.76%) |
6.69% (70.24%) |
11.63% (114.42%) |
15.45% (93.82%) |
6.29% (56.66%) |
4.02% (65.70%) |
21.93% (59.66%) |
| 三連複 | 5.87% (65.41%) |
6.47% (72.17%) |
4.87% (58.38%) |
9.30% (56.05%) |
15.45% (70.00%) |
9.27% (75.07%) |
2.87% (60.49%) |
14.35% (68.58%) |
| 三連単 | 1.26% (61.12%) |
1.75% (70.65%) |
0.81% (55.40%) |
0.00% (0.00%) |
3.64% (59.91%) |
2.65% (72.65%) |
0.29% (56.31%) |
14.35% (54.89%) |
| 総合 | 53.43% (73.26%) |
53.50% (73.51%) |
52.33% (71.05%) |
65.12% (96.02%) |
73.64% (79.71%) |
60.93% (72.58%) |
46.98% (71.90%) |
85.65% (62.53%) |
これでは障害レースは使えても他のレースでは前回公開したものに及ばないので何か対策が必要だなぁ。ML.NET 5.0.0が正式にリリースとなったのでバージョンアップと新たなモデルをと頑張ってたんですが、もう一工夫が必要なので、リリースはちょっと先になります。
0 件のコメント:
コメントを投稿