Portfolio selection is crucial for investors, but predicting stock values is tricky due to unpredictable events. The traditional model focuses on reducing risk by diversifying assets. However, when returns are not normally distributed, higher moments like skewness and kurtosis can help. A new method using a mix of mean, variance, skewness, kurtosis, and entropy was tested on real data sets. The results show that this approach is effective, especially for portfolios with higher moments.