Clusterização de dados no mercado de crédito privado: a influência das passagens entre fundos de investimento

Abstract

The main objective of this work is to analyze the impact of Transfer Procedures operations in the secondary market of fixed income for local corporate bonds, focusing on the PEJA11 debenture. For this purpose, a pattern recognition algorithm was developed using the unsupervised machine learning method DBSCAN, with the aim of identifying and differentiating these operations from effective trades. Based on a detailed analysis of the trading data registered in CETIP, covering the period from April 2023 to April 2024, it was possible to observe how Transfer Procedures operations influence the Average Daily Trading Volume (ADTV) metric and the perception of liquidity in the market. The choice of the PEJA11 debenture was motivated by its high financial volume and the significant frequency of its trades, making it representative for the study. The results obtained demonstrate that the developed algorithm was effective in identifying Transfer Procedures operations, resulting in a reduction of approximately 28% in the Average Daily Trading Volume (ADTV) after the algorithm was applied, compared to the ADTV calculated without using the method. This reduction highlights the significant influence of Transfer Procedures on the perception of liquidity. Additionally, the data visualization techniques applied provided clear and intuitive insights into trading patterns, facilitating the interpretation of the results. This study contributes to a better understanding of the nuances of the local corporate bonds market and suggests that the developed methodology can be applied to other assets, assisting in making informed decisions and more accurate pricing of these securities.