In the context of parallel collaborative lane-changing control for drivers and intelligent systems， issues such as human-machine conflicts and driving discomfort arise due to frequent or substantial changes in the control authority allocation between humans and machines. To solve these problems， this paper proposes a lateral human-machine driving weight allocation strategy that combines pre-allocation with real-time allocation to achieve a reasonable distribution of vehicle control. Initially， to characterize the driving style of the driver， a single-point preview driver model is constructed， which includes decision functions for lateral preview error and lateral acceleration. Concurrently， a model predictive controller （MPC） is established as a co-driving control system， and a vehicle lane-changing trajectory is designed based on a quintic polynomial. Subsequently， a pre-allocation method for driving rights is designed， incorporating style coefficients， preview time， and road adhesion coefficients. Criteria for real-time allocation of weights are designed based on risk level and human-machine conflict measures， with adjustments introduced to prevent frequent changes in weights. Joint simulation results indicate that when human-machine intentions are aligned， this strategy significantly reduces the driver's burden. When driving risk is high， control weights shift towards the system， allowing timely intervention to ensure traffic safety. When human-machine intentions are inconsistent， and driving risk is low but human-machine conflict is high， it is sure that control is transferred to the driver at a fixed value， allowing the vehicle to operate according to the driver's intent， and the overall control effect is superior to fixed-weight control strategies. Driver-in-the-loop platform tests show that when drivers adapt to moderate intervention by the control system， this strategy can provide personalized lane-changing assistance for drivers of different styles.