Large-Scale Benchmarking of Quantum Neural Network Configurations for Financial Time Series Forecasting
This study systematically tested 1,368 different Quantum Neural Network (QNN) configurations for predicting financial time series using GBP/USD data. The results showed that gate selection and arrangement are more important than qubit count. The best QNN configuration achieved an R2 score of 0.985, beating a classical BiLSTM model. It also highlighted that real quantum hardware noise significantly impacts performance, emphasizing the need for hardware-aware QNN design.