According to Wuhan University, Professor Luo Yong's team from the School of Computer Science has made new progress in quantum machine learning research, proving for the first time that the entanglement level of quantum data has a dual effect on the prediction error of quantum machine learning models. The relevant research results have recently been published online in the international academic journal Nature Communications. The corresponding author of the paper, Luo Yong, introduced that quantum entanglement is a key resource for achieving the advantages of quantum computing. At present, scientists are widely concerned about how to integrate quantum entanglement into various aspects of quantum machine learning models, in order to surpass the performance of traditional machine learning models. However, how the entanglement level of quantum data specifically affects the performance of quantum machine learning models remains an unresolved and challenging research topic. "Existing research generally believes that quantum entanglement helps improve the performance of quantum machine learning models." Luo Yong said that the research team analyzed the effects of quantum data entanglement level, measurement frequency, and training dataset size on the prediction error of quantum machine learning models. For the first time, it was proven that the entanglement level of quantum data has a dual effect on prediction error, which can be positive or negative. The key to determining whether quantum entanglement can improve quantum machine learning performance is the allowed number of measurements. Under sufficient measurement conditions, increasing the entanglement of quantum data can effectively reduce the prediction error of quantum machine learning models, or reduce the size of quantum data required to achieve the same prediction error. On the contrary, when the allowed number of measurements is small, using highly entangled quantum data may lead to an increase in prediction error. This study provides important theoretical guidance for designing more advanced quantum machine learning protocols, especially customized protocols for quantum computers with limited quantum computing resources. (Lai Xin She)