A Machıne Learnıng Approach For The Detectıon Of Trade Based Manıpulatıons In Borsa İstanbul

Abstract

Capital markets, one of the pillars of the financial system, play a vital role in transferring the excess funds of savers to investors who need funds in the medium and long term. Trust is essential for the safe and effective operation of capital markets and for the healthy transfer of resources from those who supply funds to those who request them. The capital market is constantly regulated and supervised in order to ensure it functions and develops in a reliable, transparent, efficient, stable, fair, and competitive environment and to protect the rights and interests of investors.

Manipulation, which prevents the capital market from operating regularly within trust and transparency norms, is an important issue that should be considered both for individual investors who offer funds to the securities markets and for companies that request funds by issuing stocks. This study examines the trade-based manipulations in Borsa Istanbul (BIST). Data on stocks that were manipulated between 2010 and 2015 were used in BIST, and a model consisting of supervised machine learning classification techniques and artificial neural networks was proposed to detect trade-based manipulation from the daily data of manipulated stocks. It has been shown that the proposed model is successful in detecting trade-based manipulations in the stock market based on accuracy, sensitivity, and f1 scores. Experimental results show that an f1 score of 0.86, a sensitivity of 0.87, and an accuracy of 0.89 in market manipulation detection were achieved.

With this study, the manipulation in the stock market, the biggest obstacle for investors to make safe investments in the capital markets, will be minimized and the principles of transparency and trust, essential for the formation and development of the capital market, will be established. In addition, due to the success achieved in market manipulation detection, our study will benefit regulators especially in detecting stock market manipulation.