Cancer, which accounts for around 24% of all fatalities that occur each year, is one of the main causes of death among people all over the world. Given that cancer originates at the cellular level, utilizing multi-omics datasets, which integrate data from genomics, transcriptomics, and proteomics, is essential for accurate detection. It is a very tedious task to detect the cancer accurately due to the huge dimensionality of the data sets. Advancement in computational techniques helps in the accurate prediction of cancer. Therefore, in the research, a Local Spatial Swarm-Guided Evolutionary Algorithm and Deep Neural Network (DNN) with optimized hyperparameters (SGE-DNN) are proposed for the accurate prediction of cancer. Initially, the multi-omics data, which consists of DNA, miRNA, and mRNA, is obtained from The Cancer Genome Portal (TCGA). Following this, the data is preprocessed to remove missing values. Additionally, genetic algorithm (GA) and local spatial particle swarm optimization (PSO) are utilized in together to carry out the feature selection. All of the features that have been gathered are combined and fed into a DNN, whose hyperparameters are adjusted via Bayesian optimization. Experiments reveal that the proposed SGE-DNN performs well with an accuracy of 98% on multi-omics datasets.