A classification framework of issuers in the Moroccan financial market

Abstract

The Moroccan financial system has undergone major changes since the early 90s. The financial market authority makes available a multitude of public information and statistics on the financial operations of the issuers. Other than the classification by sector or by type and / or amount of the issue, there is no classification model to predict the behavior of an issuer based on financial indicators. In this context, this work aims to develop an actionable classification scheme to explain and predict the behavior of issuers in the Moroccan financial market. A database of financial operations of various issuers between 1995 and 2011 was built. Thereafter, classes of these issuers were constructed via unsupervised learning techniques. Clustering of time series of issuers and their corresponding amounts reported by year, allowed for finely learning and defining classes of issuers, taking into account the temporal dimension. Based on the clusters from the first step, a supervised tree based classification model was developed to predict the class of new issuers on the Moroccan financial market.