Fill probability prediction on CAC 40 futures (test)
0.53RCLLKANFormer
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| KANFormerModel type=KAN-based Transformer, Input Features=Enriched input2025.12 | 0.53 | 0.027 | 76 | 72 | |
| DeepHitModel type=Deep Learning Survival Model, Input Features=Enriched input2025.12 | 0.56 | 0.028 | 56 | 58 | |
| ConvTrans++Model type=Convolutional Transformer, Input Features=Enriched input (agent-level info and queue position)2025.12 | 0.93 | 0.028 | 54 | 49 | |
| ConvTransModel type=Convolutional Transformer, Input Features=XLOB only2025.12 | 1.18 | 0.028 | 47 | 44 | |
| LSTMModel type=Long Short-Term Memory, Input Features=Enriched input2025.12 | 1.33 | 0.029 | 58 | 52 | |
| RFModel type=Non-deep, Input Features=Enriched input2025.12 | 1.64 | 0.027 | 64 | 63 | |
| LSTM_HazardModel type=LSTM variant, Input Features=Enriched input2025.12 | 1.64 | 0.028 | 54 | 52 | |
| CoxModel type=Non-deep, Input Features=Enriched input2025.12 | 1.74 | 0.027 | 68 | 67 | |
| MLPModel type=Multi-layer Perceptron, Input Features=Enriched input2025.12 | 1.89 | 0.029 | 53 | 51 |