Hardware-Aware Design and Layout Selection in Quantum Reservoir Computing for Insurance Claims Forecasting
Organizations: Department of Computers and Informatics, Faculty of Electrical Engineering and Informatics, Technical University of Košice, Košice, Slovakia.
Abstract
Quantum reservoir computing (QRC) processes temporal information through quantum dynamics, but it remains unclear whether small QRC models can forecast real data competitively with classical models and whether their predictions survive the transfer from ideal simulation to noisy hardware. We address both questions for a six-qubit gate-based reservoir that forecasts next-day motor-insurance claim counts. Five reservoir candidates were selected and frozen using hardware-native interaction graphs, Trotterized Ising dynamics, memory and measurement-robustness criteria, circuit cost, and calibrated-noise screening. On 232 held-out test days, the ideal candidates reached root-mean-square errors (RMSEs) of 5.66-6.12 claims/day, compared with 5.36 for ridge regression. For four of the five, the difference from ridge was not significant under a 95% block bootstrap, and their accuracy was comparable with classical recurrent and reservoir baselines. On the 156-qubit IBM Kingston processor, we ran 63 evaluations in 21 paired rounds, comparing hardware-aware, random-native, and Qiskit optimization level 1 layouts. Hardware-aware placement gave lower RMSE than Qiskit in every round, with a maximum transfer distortion of 7.59 claims/day (11.84 for Qiskit), and avoided the severe transfer failures seen for some random layouts. Most of the hardware distortion appeared as a systematic forecast shift, which a short hardware probe on training inputs detected before the test runs. Longer replay windows raised circuit cost without a consistent gain in forecasting accuracy or hardware transfer. Small QRC models can thus reach competitive accuracy on this task, and hardware-aware layout selection makes their transfer to IBM Kingston more consistent.
Figures & tables
| Work | Encoding | Reservoir; readout | Hardware (qubits) | Task |
| Fujii and Nakajima [ 6 ] | State injection | Transverse-field Ising; linear | Numerical simulation ( – ) | Temporal benchmarks |
| Kutvonen et al. [ 13 ] | State injection | Fully connected transverse-field Ising; linear | Numerical simulation ( ) | Stock prices |
| Chen et al. [ 8 ] | Gate encoding | Gate-based; linear | IBM Boeblingen/Ourense/Vigo ( – ) | Temporal benchmarks |
| Suzuki et al. [ 14 ] | Gate encoding | Natural noisy dynamics; linear | IBM Melbourne/Toronto ( – ) | NARMA, classification |
| Wudarski et al. [ 15 ] | Projection + angles | Fixed entangling; ridge | OQC Lucy ( – ) | Chaotic systems |
| Hamhoum et al. [ 16 ] | Angle encoding | Trotterized transverse-field Ising; ridge | IBM Heron R2 ( – ) | Lorenz-63, ENSO |
| Input | Pearson | MI ∗ ( ) | Main evidence |
| – | ( ) | 24% of target variance explained by weekday | |
| ( ) | beyond weekday | ||
| ( ) | Nonlinear dependence despite weak | ||
| ( ) | beyond calendar | ||
| Note: ∗ MI is measured in nats; is obtained from target permutations. denotes the additional explained variance after adding the predictor. Pearson is omitted for weekday because its numerical labels represent categorical values. | |||
| Frozen configuration | Ising couplings | |||||||||||||||||||
| Graph | CV | Val. | CZ | ( s) | ||||||||||||||||
| C1 | H3 | – | – | |||||||||||||||||
| C2 | H3 | – | – | |||||||||||||||||
| C3 | H2 | – | – | |||||||||||||||||
| C4 | H0 | – | – | |||||||||||||||||
| C5 | H2 | – | – | |||||||||||||||||
| Rounds | Measurement settings | Observables | Qiskit layout | |
| C1 | 5 | XXXZYX , ZZZZZZ | , , | VF2Layout |
| C2 | 5 | XXXZ , ZZZZ | , | VF2Layout |
| C3 | 5 | XXXZ , ZZZZ | , | VF2Layout |
| C4 | 3 | XXXXYX , ZZZZZZ | , , | Trivial |
| C5 | 3 | XXXX , ZZZZ | , | VF2Layout |
| Notes: Measurement settings specify the measurement bases of – ; qubits not listed are not measured. Qubits – encode , , , and , respectively, while and constitute the memory register. The readout is omitted from C1–C3 based on the R3 candidate-selection criterion. Each measurement setting uses 2048 shots. | ||||
| Model | RMSE | MAE | Bias | cRMSE | vs ridge | vs XGBoost (sq.) |
| ridge | — | [ , ] | ||||
| NB2 GLM | [ , ] | [ , ] | ||||
| ESN † | [ , ] | [ , ] | ||||
| GRU † | [ , ] | [ , ] | ||||
| vanilla RNN † | [ , ] | [ , ] | ||||
| XGBoost (squared) | [ , ] | — |
| Period | Mean | Clipped (%) | Policies ( ) | C per pol. | ||
| training | – | – | ||||
| validation | – | – | ||||
| test 1. half | – | – | ||||
| test 2. half | – | – |
| Candidate | Policy | Rounds | RMSE | Bias | Median | Max. | Median |
| C1 (ideal ) | HA | ||||||
| Random | |||||||
| Qiskit | |||||||
| C2 (ideal ) | HA | ||||||
| Random | |||||||
| Qiskit |
| Ideal RMSE | HA RMSE | HA | Circuit s | ||||
| C1 | ( ) | ||||||
| C2 | ( ) | ||||||
| C3 | ( ) | ||||||
| C4 | ( ) | ||||||
| C5 | ( ) | ||||||
| : Rule-R2 memory score, mean (SD) over five seeds. : effective memory horizon in days based on the -percentile permutation threshold; Ideal RMSE & HA RMSE: test RMSE; HA : median transfer distortion; circuit duration is the scheduled duration of the longer measurement setting. | |||||||
| Method | Rank | CV | Val. | Sel. | Par. | Configuration |
| Baseline | – | – | 0 | seasonal_naive_7d | ||
| Baseline | – | – | training_mean | |||
| Baseline | – | – | persistence | |||
| ridge | alpha= | |||||
| ridge | alpha= | |||||
| ridge | alpha= |
| Hardware-aware | Random native | Qiskit | ||||||||||||
| Ideal | Aer | QPU | Aer | QPU | Aer | QPU | ||||||||
| Candidate | Round | RMSE | RMSE | RMSE | RMSE | RMSE | RMSE | RMSE | ||||||
| C1 | ||||||||||||||
| C1 | ||||||||||||||
| C1 | ||||||||||||||
| C1 | ||||||||||||||
Quantum Reservoir Computing
Quantum reservoir computing uses quantum dynamics to transform time-dependent inputs into features for prediction and classification.
See all 20 papers