Financial Product Recommendation through Case-based Reasoning and Diversification Techniques

This work proposes a framework for financial product recommendation which combines case-based reasoning with diversification techniques to support wealth managers in recommending personalized investment portfolios. The performance of the framework has been evaluated against 1172 real users, and results show that the yield obtained by recommended portfolios overcomes that of portfolios proposed by human advisors in many experimental settings. 1. MOTIVATIONS AND METHODOLOGY Widespread recommendation approaches, such as contentbased (CB) and collaborative filtering (CF), can hardly put into practice in the domain of financial recommendations. Typically, pure CB strategies are likely to fail since content information describing both users and financial products is too poor to feed a CB recommendation algorithm. Moreover, the over-specialization problem, typical of CB recommenders, collide with the fact that turbulence and fluctuations in financial markets suggest to change and diversify the investments over time. On the other side, CF algorithms may lead to the problem of flocking, since user-based CF could move many (similar) users to invest in the same asset classes at the same time, making the algorithm victim of potential trader attacks. As a consequence, we focused the attention on different recommendation paradigms. Given that financial advisors have to analyze and sift through several investment portfolios before providing the user with a solution able to meet his investment goals, the insight behind our recommendation This work fullfils the research objectives of the projects ObjectWay-Finance-as-a-Service: Smart Application software and Service for Financial Services Operators and PON 01 00850 ASK-Health (Advanced System for the interpretation and sharing of knowledge in health care). http://www.technologyreview.com/view/425654/flockingbehaviour-improves-performance-of-financial-traders/ http://en.m.wikipedia.org/wiki/Portfolio (finance) Copyright is held by the author/owner(s). RecSys 2014 Poster Proceedings, October 6-10, 2014, Foster City, Silicon Valley, USA. framework is to exploit case-based reasoning (CBR) to tailor investment proposals on the ground of a case base of previously proposed investments. Formally, given a case library C, each case ci ∈ C is a triple 〈ui, pi, fi〉, where ui is a representation of a user, pi is a representation of the portfolio, and fi is a feedback assessment. Each user ui is represented as a vector of five features: risk profile, inferred through the standard MiFiD questionnaire, investment goals, temporal goals, financial experience and financial situation. Our recommendation process is based on the typical CBR workflow, described in [1], and is stuctured in three different steps: (1) Retrieve and Reuse: retrieval of similar portfolios is performed by representing each user as a vector according to the weight of each feature (very low=1, very high=5). Next, cosine similarity is adopted to retrieve the most similar users (along with the portfolios they agreed) from the case base. (2) Revise: the candidate solutions retrieved by the first step are typically too many to be consulted by a human advisor. Thus, the Revise step further filters this set to obtain the final solutions. Five techniques have been introduced for the revise of the list: (a) Basic Ranking: portfolios are ranked in descending cosine similarity order, according to the scores returned by the Retrieve step. The first k portfolios are returned to the advisor as final solutions. (b) Greedy Diversification: this strategy implements the diversification algorithm described in [3]. The algorithm tries to diversify the final solutions by iteratively picking from the original set of candidate solutions the ones with the best compromise between cosine similarity and intralist diversity with respect to the previously picked solutions. At each step of the strategy, the solution with the best compromise is removed from the set of candidate solutions and is stored in the set of final solutions. (c) FCV: this strategy adapts the Interest Confidence Value proposed in [2] to the financial domain. Financial Confidence Value (FCV) calculates how close to the optimal one is the distribution of the asset classes in a portofolio, according to the average historical yield obtained by each class. Given a set of asset classes A, for each portfolio p the set P , which contains the asset classes which compose it, and its complement P are computed. Next, FCV is formally defined as: