Abstract: In people’s daily life, there are plenty of decisions need to be made among various options, and it is usual to choose an option by balancing risk with profit. To solve this kind of problem, existing recommender systems rely on some relevant data or history data of user or similar users. So a recommendation method which can make recommendations by balancing risk with profit and doesn’t rely on users’ data needs to be proposed in the situation where history data is lack or the history data is hardly to use such as recommending places for passenger to find vacant taxi. In this paper, a recommendation method is proposed which can be widely used in above situation based on the Modern Portfolio Theory（MPT）. Then, this method is used in recommending places for passenger to find vacant taxi to be an example and shows how to choose the more proper calculating method of risk and profit compared with other work which uses MPT to make recommendations. This method is tested on real world dataset to verify the balance performance.

Key words: recommendation method, portfolio, risk and profit, probability distribution

摘要： 日常生活中，人们面临众多需要在可选对象中进行抉择的问题。其中一些往往需要衡量风险及收益进行决策，对此已有的推荐方法依赖于用户或相似用户的历史数据，因此在类似打车地点推荐等缺乏这些数据或类似数据可重复利用度低的情况下，需要一种不依赖用户方面数据，同时能够权衡可选对象的风险及收益进行推荐的推荐方法。以经济学领域的现代投资组合理论为基础，提出一种可应用于上述场景的推荐方法。并以打车位置推荐为例说明如何使用该方法，以及同以往对于该理论的应用相比，应该如何更为适当地选择风险及收益的计算策略。在真实的数据集上进行实验，验证方法中权衡推荐策略的有效性。

关键词: 推荐方法, 投资组合, 风险与收益, 概率分布