OPTIMAL PORTFOLIO FORMATION USING MEAN VARIANCE EFFICIENT PORTFOLIO AND CAPITAL ASSET PRICING MODEL WITH ARTIFICIAL NEURAL NETWORK AS STOCK SELECTION METHOD

There are two main things in forming an optimal stock portfolio: stock selection and stock weight determination. This study aims to determine the performance of an optimal portfolio formed using ANN as a stock selection method and MVEP (Mean-Variance Efficient Portfolio) and CAPM (Capital Asset Pricing Model) to determine stock weights. In addition, it is also necessary to determine the characteristics of the stocks formed in the portfolio. The criteria for stock selection are choosing stocks predicted to have maximum mean returns with minimal risk. This research uses data from 10 stocks listed on the Indonesian Stock Exchange. The forecasting results state that ANN can be used to predict stock prices to get a picture of stock prices in the future. Based on the calculation results, BMRI, TLKM, ASII, TPIA, and BBNI stocks were selected to form a stock portfolio. The MVEP and CAPM methods produce stock weights with different characteristics. The MVEP method gives the most significant weight to stocks that have the largest predicted mean return but experience changes in accuracy categories. The CAPM method gives the most significant weight to stocks with less risk than other stocks and has the smallest MAPE value. Empirically, ANN can be used to select stocks to form a portfolio. Stock price predictions with the most significant mean return and small risk can be used as a reference when forming a portfolio using the MVEP and CAPM methods.