Capital Asset Pricing Model and Momentum Ensembled Strategy based Portfolio Design by Genetic Algorithm

Abstract

Designing a successful investment strategy which yields greater profits while avoiding risks is a very challenging task due to unpredictability of future events in the financial market. Moreover, investors cannot ceaselessly make correct investment decisions because countless many socio-economic factors often threat the stability of the fiscal products. Earlier studies in finance suggested that constructing a portfolio, that promises risk-spread gains, can offer a good starting point. Later researchers improved the traditional ideas on portfolio while reducing the computational complexity and articulating the associated relationship among stocks in the portfolio better. Inspired by the earlier works, yet pointing out their insufficiency as a winning investment initiative, we attempt to design an optimal portfolio comprised of outperforming items in terms of returns per risk with potentials to grow further. While many existing machine learning methods in the finance seek for the future price, we allocate stocks that have high potentials. More specifically, we first analyze their risk-adjusted returns in the previous term and use their inertia as a momentum strategy. However, because historic price movements alone cannot fully explain their future changes nor guarantee positive returns, we recognize the needs for more informed estimations in investment decisions. Using Capital Asset Pricing Model(CAPM), we calculate the values of each stock and determine whether it is under or overvalued. In this study, we used a Genetic Algorithm to design the CAPM and Momentum ensembled investment plans. Such approach designed by a Genetic Algorithm was tested in four separate stock markets and was verified, not just as an analytic tool of past behaviors but as a future investment platform.