Recommender systems in the domain of early-stage enterprise investment : investment decision-making & venture valuation

Abstract

The main objective of this thesis consists of the design of a recommender system, representing a novel method concerning the computational recommendation of early-stage enterprises to investors. In order to quantify decision rules the recommender system is based on, investors requirements and behaviours need to be analysed utilizing qualitative- and quantitative research. Furthermore, demonstrating the behaviour of the proposed recommendation algorithms is a major task of this thesis. For this reason, a prototype of the recommender system is being crafted in software. Due to the fact that the usability of the recommendation systems user interface plays a key role in terms of recommendation quality, a usability- and recommendation quality review of the prototype is being conducted in the course of empirical research. Based on the results of the Investment Decision-making & Venture Valuation specialization topic, it can be concluded that the most important characteristics investors base their investment decisions on, are stated as the quality, size and composition of the management team, product- & public interest and the industry / market sector of an early-stage enterprise. Furthermore, the venture valuation methods most utilized by investors, most meaningful in terms of valuation quality in the context of early-stage enterprises and most beneficial when utilized in a recommendation system, are stated as the scorecard- and berkus methods. Finally, investors requirements among the functionality of a recommender system in the domain of early-stage enterprise investment may be concluded as the construction of an investor profle. The shared chapter 3 Recommender Systems for Early-Stage Enterprise Investment addresses the conceptualization of a recommendation system in the domain of early-stage enterprise investment based on the findings of co-author Christian Ohrfandls specialization topic Investment Decision-making & Venture Valuation. The resulting recommender system includes various types of recommenders in a parallelized approach, that is, Collaborative filtering, content-based-, knowledge-based-, social- and hybrid recommendation algorithms. Additionally, the conceptual model of this recommender system has been implemented as a highly scalable, plugin-based software prototype that may be easily extended by different recommendation algorithms in future work. The most important opportunity for future research is stated as qualitative- or quantitative evaluations of recommendation quality in terms of user satisfaction. These evaluations may answer the question, whether the implemented design decisions improve a users utility when using the system. In fact, it is precisely this very evaluation that is being researched by co-author Johannes Luef in the course of the specialization topic User-centred Evaluation.