Towards Building Ranking Models with Annual Reports

AbstrAct: The textual content of company annual reports has proven to contain predictive indicators for the company fu­ ture performance. This paper addresses the general re search question of evaluating the effectiveness of applying machine learning and text mining techniques to building predictive mod­ els with annual reports. More specifically, we focus on these two questions: 1) the feasibility of build ing ranking models with annual reports to rank future firm performance and 2) the ef­ fect of integrating meta seman tic features to help improve and support our prediction. We compare models built with different ranking algorithms and document models. We evaluate our models with a simulated portfolio. Our results show significantly positive average returns over 5 years with a power law trend as we increase the ranking threshold. Adding meta features to document model has shown to improve ranking performance. The SVR & Meta­augemented model outperforms the others and pro vides potential for explaining the textual factors behind the prediction.