Nowadays swarm intelligence algorithms are becoming increasingly popular in solving many optimization problems. But different algorithms have their own strengths and weaknesses according to their composite designs. Wolf Search Algorithm (WSA) is a contemporary semi-swarm intelligence algorithm designed to solve complex optimization problems, and demonstrated its capability in local intensified search and global exploration search, especially for large-scale problems. However, it still inherits a common weakness just like other swarm intelligence algorithms. Its performance is heavily dependent on the chosen values of the control parameters. In 2016, I proposed and published Self-Adaptive Wolf Search Algorithm (SAWSA) which offers a sim-ple solution to the parameters adaption problem. As a very simple schema, the original SAWSA parameters adaption is based on random guesses, which is unstable and naive. Based on the SAWSA, I investigate the WSA search behavior more deeply. A new parameter guided updater, which has a Gaussian guided parameter control mechanism based on infor-mation entropy theory, is proposed as an enhancement of the SAWSA. The heuristic updating function is improved. The new method denoted as Gaussian Guided Self- Adaptive Wolf Search Algorithm (GSAWSA). From the experience of GASAWSA enhancement, I have observed and obtained knowledge that metaheuristic algorithms carry their own unique functions which are implemented to mimic certain search patterns or beahviours. In computer science, inventing new swarm search algorithm has emerged through a booming era. Many new efficient metaheuristic algorithms were invented and published. However, most of them are similar in principle designs, different in variation of certain functions in their search behaviours. Given the sheer number and variety of variants, it is not practically possible to compare all the exist metaheuristic algorithms in order to find the very best solution for a specific optimization problem. Each algorithm, old or new, all have their own scenarios to be the most fitting algorithm. This phenomenon is known as No Free Lunch Theorem, which is a theoretical finding that suggests all optimization algorithms perform equally well when their performance is averaged over all possible objective functions. Some algorithms at some points of run time will exhibit their peak optimization performance under running some application problems. In my research study, instead of following the flow to make some slight changes, rebranding it and claiming credits of a new invention, I looked into the common and unique characteristics of metaheuristic algorithms, putting them over a benchmarking platform. By monitoring and interchanging the composite functions of the selected classical metaheuristic algorithms under some standard bench- mark functions (or real industry problem) during different stages of a searching process, good final solutions are found. By this approach, a Boosting (piggyback) system and a Stacking (carry forward) large scale ensemble system are proposed. The systems try to make use of the best aspects of existing algorithms, and collectively improve the results progressively over a sequence of checkpoints. By following the mix-and-match concept, I invented another solution to combine several classic swarm intelligence algorithms using a brick-up recombination model. Brick-up is analogous to building up a composite optimization solution using the best parts or components of different algorithms, like Lego bricks. Our proposed model helps users to construct new composite algorithms from a collection of components from the most suitable metaheuristic optimisation algorithm candidates without any professional knowledge. This thesis contains all the research steps from the WSA to a large scale ensemble system and a brick-up model for metaheuristic algorithm. All the proposed algorithms and models are tested with simulation experiments using benchmarking functions as well as real-life datasets.