Spectral unmixing (SU) is used for the identification of the end members and fractional abundances present in the mixed pixels of the hyperspectral images that are rich in data. The traditional methods for spectral unmixing are affected with the artifacts requiring the effective method for spectral unmixing. Thus, the paper uses an effective framework for Spectral unmixing using the proposed Rider based Elephant Herd Optimization (R-EHO). The proposed Rider based Elephant Herd Optimization estimates the fractional abundance using the end members and proposed Rider based Elephant Herd Optimization is the integration of the standard Elephant Herd optimization (EHO) in the Rider Optimization algorithm (ROA). The proposed optimization uses the multi-objective function that is modelled using the factors, such as Reconstruction Error (RSE), sparsity (SPA), spatial neighbor (SNI), and spatial neighbour correlation (SNC), and mutual information. The analysis is progressed using the Urban and Cuprite datasets based on the performance metrics, Root-Mean-Square Error (RMSE), Reconstruction Error (RE) and it is proved that the proposed R-EHO acquired the minimal RE and minimal RMSE of 13.8576 and 0.0011, respectively.