This thesis proposed a new method namely Hybrid-EDAfold which is an evolutionary algorithm (EA) based on a hybrid estimation of distribution algorithms (EDAs) for RNA secondary structure prediction. The proposed method consists of two EDAs and using minimum free energy technique. The Hybrid-EDAfold uses both good and poor solutions enabling the algorithm to search throughout the search space. Using information from poor solutions can indicate which area is unappealing to explore when conducting a search with high-dimensional data. In addition, one of the EDA uses a mutation operator to support local search which increases the diversity and moderately avoid early convergence. Moreover, the proposed method returns the answer as a set of structures consisting of optimal structure and suboptimal structures to increase the chance of finding a predicted structure closer to the real structure. Comparison of the Hybrid-EDAfold was evaluated with well-known web servers namely Mfold, RNAfold, and RNAstructure on 15 RNA types with 760 RNA sequences total. The Hybrid-EDAfold yields better results than other methods in every metrics. The proposed method was also compared with metaheuristic methods on 20 RNA sequences collected from their literature. The results showed that the Hybrid-EDAfold yields better results than RnaPredict and SARNA-Predict and is comparable to TL-PSOfold.
