Predicting stock trends through technical indicators is currently a highly popular research area. It primarily involves generating buy and sell signals through technical indicators to form a trading strategy portfolio (TSP), with the goal of increasing returns while mitigating risks. In the previous approach, an algorithm was designed to find a TSP based on the mean-semivariance model (MSV model) using the memetic algorithm (MA). However, in practice, numerous uncertainties may affect the profitability and hedging ability of trading strategies. To handle this problem, based on the fuzzy mean-semivariance model (FMSV model), an optimization algorithm is proposed to find a TSP using the memetic algorithm. In the global search, the genetic algorithm (GA) is employed. It first uses the selected technical indicators to generate candidate trading strategies. Then, the trading strategies are used to generate TSPs as chromosomes. Every trading strategy in a chromosome is represented by three parts: weight, buy and sell indicators, and parameters of indicators. The fuzzy return, risk, and trading frequency of a chromosome are calculated to evaluate fitness value of every chromosome. Next, in the local search, the simulated annealing (SA) is used to tune the parameters of the used technical indicators in chromosomes. Then, the genetic operators are applied to the population to form new offspring. The evolution process is repeated until the stop criteria are reached. Experimental results were made on real datasets and indicate that the proposed approach is capable of maintaining stable returns and mitigating risks compared to the existing methods.