分かってみればあれですが、クラスのコンストラクタの記述でコピペしてて、コピペ後に修正が必要なのにしてなかったのが原因でした。順位予測のこれまでの最適RMSE 3.7163は
| 1,031R | 1点 | 芝(541R) | ダート(452R) | 障害(38R) | 8頭以下(93R) | 9～12頭(323R) | 13頭以上(615R) | 多点 |
| 単勝 | 25.02% (82.64%) |
22.74% (77.86%) |
27.21% (89.09%) |
31.58% (73.95%) |
37.63% (108.28%) |
29.72% (82.29%) |
20.65% (78.94%) |
52.57% (77.47%) |
| 複勝 | 55.19% (85.60%) |
51.57% (80.91%) |
59.29% (92.08%) |
57.89% (75.26%) |
69.89% (90.75%) |
61.61% (88.45%) |
49.59% (83.32%) |
87.49% (80.90%) |
| 枠連 | 13.48% (90.70%) |
12.82% (92.89%) |
13.81% (88.29%) |
21.05% (94.74%) |
-- (--) |
14.55% (85.05%) |
11.38% (83.35%) |
25.23% (80.12%) |
| 馬連 | 9.70% (81.51%) |
9.80% (87.32%) |
9.29% (74.82%) |
13.16% (78.42%) |
12.90% (32.15%) |
12.07% (83.37%) |
7.97% (88.00%) |
21.05% (77.47%) |
| ワイド | 22.79% (85.84%) |
20.89% (73.31%) |
24.56% (100.66%) |
28.95% (87.89%) |
39.78% (69.68%) |
26.32% (79.91%) |
18.37% (91.40%) |
42.77% (75.12%) |
| 馬単 | 5.53% (82.21%) |
4.62% (81.24%) |
5.97% (80.20%) |
13.16% (120.00%) |
6.45% (21.72%) |
7.12% (70.59%) |
4.55% (97.46%) |
21.05% (73.33%) |
| 三連複 | 5.04% (54.80%) |
4.99% (49.09%) |
4.65% (60.04%) |
10.53% (73.68%) |
16.13% (52.15%) |
8.05% (69.72%) |
1.79% (47.37%) |
13.39% (71.10%) |
| 三連単 | 0.78% (43.60%) |
0.18% (0.92%) |
1.11% (82.63%) |
5.26% (186.84%) |
1.08% (5.38%) |
1.55% (37.62%) |
0.33% (52.52%) |
13.39% (73.79%) |
| 総合 | 56.45% (75.56%) |
53.05% (67.28%) |
60.40% (83.43%) |
57.89% (99.12%) |
72.04% (54.30%) |
62.23% (74.62%) |
51.06% (77.79%) |
87.97% (74.80%) |
こんな感じでした。まだまだ精度が上がらないと使えないですね。それと、まだ完全な状態ではないものの、アプリに添付するのかも悩む所です。
0 件のコメント:
コメントを投稿