基于类比推理的短生命周期无形体验品需求预测

Development of intangible experience goods with short life cycle has been a national strategy for more than ten years, but many of these goods are difficult to recover their cost because of lacking effective demand forecasting. The existing demand forecasting techniques and methods, mainly developed for tangible functional goods with long and short life cycle, are not adapted to intangible experience goods with short life cycle. Aiming at addressing the difficulty to forecast intangible experience goods with short life cycle resulted from little or no sales data, this project will propose a series of models and methods for forecasting the goods by taking the goods as analogy target and movie as analogy source, with demand forecasting based on analogy as the core, guided by practical schema theory, structure mapping theory and other analogical reasoning theories, by researching overall surface feature models of movie products and the goods, their mapping relationships of overall surface features between movie products and the goods, practical schema of movie box office, methods for retrieving movie analogy, forecasting methods by analogical transferring from movie analogy source to the goods analogy target, and other key scientific issues. Furthermore, this project will verify the models and methods by applying them to demand forecasting of mobile games. The verified models and methods will provide theoretical and technical supports for effectively forecasting demand of the goods, and ultimately achieve a comprehensive upgrade of production and marketing investment success rate of the goods in China.

十多年来，发展短生命周期无形体验品已成为国家战略，但因欠缺有效需求预测致使众多该类产品难以收回成本。现有的需求预测技术和方法主要针对长生命周期和短生命周期有形功能品提出的，难以适用于短生命周期无形体验品。针对短生命周期无形体验品因预测对象没有或仅有少量历史销售数据导致的需求预测难问题，本项目以该类产品为类比靶，以电影为类比源，以基于类比推理的该类产品需求预测为核心，以实用图式理论、结构映射理论等类比推理理论为指导，系统研究电影总体和该类产品总体的表面特征构建、两者总体表面特征间的映射关系、电影表面特征与票房间关系的实用图式归纳、相似电影源的检索、从电影源到该类产品的类比迁移预测、实例化刻画等关键科学问题，以手游为典型案例进行理论验证和应用，面向该类产品提出一套基于类比推理的需求预测模型和方法，为有效预测该类产品需求提供理论和技术支撑，最终达到全面提升我国该类产品生产和营销投资成功率的目的。

因欠缺有效需求预测致使众多短生命周期无形体验品难以收回成本。以往的需求预测技术和方法主要针对长生命周期和短生命周期有形功能品，难以适用于短生命周期无形体验品。针对短生命周期无形体验品因预测对象没有或仅有少量历史销售数据导致的需求预测难问题，本项目进行了五方面研究。（1）产品需求特征模式挖掘研究，得到了四种适用于新产品开发前使用、源于不同信息的需求特征模式挖掘方法。（2）网络社交平台上新产品信息扩散预测研究，得到了三个考虑个体层面特点的群体层面网络社交平台上产品信息扩散模型。（3）新产品扩散机理研究，得到了三个考虑消费者异质性的新产品扩散动力学机理模型，包括新产品消费者涌现行为数理模型、耦合改进SEIR流行病模型与异质主体模型的非线性动力学模型、考虑产品信息发布方式影响的新产品扩散耦合模型，以及三个描述产品信息与新产品扩散之间的复杂因果关系模型。（4）产品上市前需求总量预测研究，得到了7种预测方法，包括基于改进证据理论的短生命周期新品上市前总销量预测方法“多证据体动态加权融合预测方法”、电影总票房预测的偏最小二乘回归模型、结合粗糙集与证据理论的区间信度分类预测方法、结合区间理论和熵权法及理想解法的映前总票房区间预测方法、基于改进BP_Adaboost算法的首轮融资时电影总票房分类预测方法、融合电影海报特征等多模态信息的映前总票房分类预测模型。（5）以电影为例的产品分时段需求预测研究，得到了6种预测方法，包括考虑消费动机和动态竞争的电影日需求预测模型、融合Bass模型和三阶段过程模型的续集电影需求扩散模型、结合改进Bass模型和参数回归模型的电影日需求预测模型、基于类比法的短生命周期产品需求扩散预测方法、考虑竞争的电影首映日票房集成预测方法、基于电影预告片特征的电影首映周票房收入预测模型。研究成果为有效预测短生命周期无形体验品需求提供了理论和技术支撑，可为我国该类产品生产和营销投资提供决策支持。