【摘要】 手机信令数据在时空行为研究中得到了广泛应用，但由于缺乏活动目的的语义信息，限制了其在城市规划中的应用潜力。为了解决这一问题，论文提出一种改进方法，用于推断手机信令数据中非就业活动目的。该方法融合了居民交通出行调查数据和活动地到访频率等多源数据，并采用多项logit模型探究活动目的与个体属性信息、活动时间特征和活动地空间属性之间的关联规律。此外，论文创新性地引入了反映长周期行为模式的到访地频率变量。实证结果表明，加入到访频率后，模型拟合优度从0.265提高至0.442，整体预测准确率从58.0%升至69.2%。这种方法在保持模型解释性的同时，提升了非就业活动识别的准确率，为深入理解居民时空行为需求提供了新的途径。研究成果可以为公共设施规划、交通需求预测和商业布局优化等领域提供有力的数据支持和决策参考。更多还原

【Abstract】 Mobile phone signaling data have been widely used in spatiotemporal behavior research, but their potential application in urban planning is limited due to the lack of information on activity purposes. To address this issue, this study proposed a method to infer the purposes of non-work activities from mobile phone signaling data. The method integrates multisource data, and employs a multinomial logit model to explore the relationship between activity purposes and individual attributes, temporal characteristics, and spatial properties of activities.The key innovation lies in the introduction of a location visit frequency variable that reflects long-term behavioral patterns. Empirical results show that incorporating visit frequency significantly improves the model’s goodness of fit from 0.265 to 0.442, and increases the overall prediction accuracy from 58.0% to 69.2%. While maintaining interpretability, this method substantially enhances the accuracy of non-work activity identification,providing new insights into residents’ spatiotemporal behavior patterns. The findings offer valuable data support and decision-making references for public facility planning, traffic demand forecasting, and business spatial layout optimization.更多还原