The recommender system is the most developing application in the web environment. A good Recommendation system needs to make choices and decisions to provide recommendations to the user. The Recommender system filters the information and provides accurate recommendations of products or services for a concerned individual. In this research, we propose a model, which employs three approaches; clustering, nature-inspired optimizer, and deep learning. The fuzzy c-means clustering technique is combined with the particle swarm optimization algorithm. The particle swarm optimizer works as a feature extractor. The particle swarm optimizer is applied to the Movielens dataset to get the initial cluster positions. The fuzzy c-means is used to classify the users in the dataset and to reduce data redundancy. The optimized and clustered data is provided as input to the autoencoder to get the final recommendations of the movies to the users. We have proposed our recommender system in two parts; 1) Using a single-layer autoencoder, and 2) Using a deep autoencoder. We analyzed our proposed system over a publicly available dataset- Movielens. The proposed system is compared with the existing recommender systems. The experimental results of our system revealed that the performance and efficiency are improved, and offered better recommendations.