Alzheimers disease (AD) is an incurable neurodegenerative disease that mainly affects the aged population. For early AD diagnosis, there is a requirement for effective automated techniques. Researchers have propounded numerous novel approaches for AD classification. Nevertheless, for more comprehensive knowledge about AD research, more effective learning techniques are essential. Application of feature selection has a major impact on the classification procedure's speed due to the removal of unnecessary features. The feature selects an optimized feature subset from a larger feature set. This work has proposed the feature selection utilizing a hybrid Krill Herd-Genetic Algorithm (KH-GA) algorithm and optimizes the Neural Network (NN). The computation of the NN's weights or optimization minimizes the function cost or error for attaining feasible outcome. The stochastic nature-inspired optimization algorithm known as the Krill Herd (KH) has been successfully applied to resolve numerous complex optimization problems. The KH algorithm's performance can sporadically get influenced by its poor exploration (diversification) ability. To attain the best global optima, the KH-GA algorithm boosts the KH algorithm's global (diversification) search ability. This proposed algorithm has been operated via the addition of the Genetic Algorithm's (GA) global search operator. This will boost the KH'sexploration around the optimal solution and kill individuals that shift towards the global best solution. In comparison to the other approaches, the experimental outcomes have demonstrated the proposed algorithm's superior performance.